
From nobody Sat Jan  1 23:56:09 2022
Return-Path: <do_not_reply@mnot.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id EE4DC3A1F71 for <ppm@ietfa.amsl.com>; Sat,  1 Jan 2022 23:56:00 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=mnot.net header.b=MBfDX768; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=jGtw3YA2
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id YHWSudDFz08I for <ppm@ietfa.amsl.com>; Sat,  1 Jan 2022 23:55:55 -0800 (PST)
Received: from wout3-smtp.messagingengine.com (wout3-smtp.messagingengine.com [64.147.123.19]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 4D6D03A1F72 for <ppm@ietf.org>; Sat,  1 Jan 2022 23:55:55 -0800 (PST)
Received: from compute4.internal (compute4.nyi.internal [10.202.2.44]) by mailout.west.internal (Postfix) with ESMTP id 878323200EC0 for <ppm@ietf.org>; Sun,  2 Jan 2022 02:39:56 -0500 (EST)
Received: from mailfrontend2 ([10.202.2.163]) by compute4.internal (MEProxy); Sun, 02 Jan 2022 02:39:56 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mnot.net; h= content-type:mime-version:from:to:subject; s=fm1; bh=l4nJ2B3dBQU v02Iu+Hw4GZD1iyW8tH3zGQzmez0Og9U=; b=MBfDX7688L6sNhxCL6vxJHb9Z85 R8t3bET6VJ9SyNRGZan11qPKRrQiZYNTJ08idj+Tk738IRC0VOANRfcQotInX3zH WrV8BjoJjitHtf/vwfPA+5+yMuHhxyF7A6yCiXX4KfmQpwzK6Bov63DHwnhjMJo0 p4wwNfeLyXivdF616gclh6qxhFkB+9Ybk4gPAEn2ehGPThbVm3BR1GJicLjrR2H8 OoPQFlUv7Xap8/r1LzgSxM2iR1dvZZqonfPHrRfkWx/mvW+7iQgw+hV1q2879jOc QIw3uJGS8AmfeZY3Ni84KIndTPlDAyR8xnHwr4OAABNq9AKtEXjR38Tqcnw==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=content-type:from:mime-version:subject:to :x-me-proxy:x-me-proxy:x-me-sender:x-me-sender:x-sasl-enc; s= fm1; bh=l4nJ2B3dBQUv02Iu+Hw4GZD1iyW8tH3zGQzmez0Og9U=; b=jGtw3YA2 Ml5ISJ5H8LWS62YhaM7PcU5D6eGFiRKfINo79v50GyA96fGYocwOh1rUBpthfdj6 2zd1FLS4q1uBZB7AfiIb37cTitk4f/0PFXZoNCT/EahyIcGMxSyMp5iVRDxyW1wv sHDsgbFdDspg40qxaSJ7h7dzjJ3bvn4dtnMrnJEPJtaGbM4RKTUE5YWDud73bgmm uaw9rilhA7wgrWpfXsvob/kkr9KEj4QZvG/1Ze3jv3ZoTDDQVXWd6gODrkZ67BQ5 Usx9V9RT56/jfm/9DiERWqjGSUYiFf+TbUDgRVDHVNuZOBiNkRt45UkWT2WO37al EHVWfoPRz0bz3w==
X-ME-Sender: <xms:y1bRYSPLk_7GnJ38mLD7Mq41_V4MVMDPBhL_883UQShyoEr_rhVkPw> <xme:y1bRYQ8MqRPorjUf325__l6UNDCN7QfGXzGzo0ODSbkCJ6H2TJenxEJ-ZnOSr6dmW gOgqTE8-iSEHAlUbw>
X-ME-Received: <xmr:y1bRYZT0A1eWo_cdSGUWbkD5rZothzbtyl316aI1PCCmZHO8DW4sFC1LyLdlI4Byqna2o_uMctBa4DSw8moZ3UKtXeLYYcIHvm_iSBijEqvXO0aCgBhgkD1Ui_SaIhWMqvc>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvuddruddvkedguddutdcutefuodetggdotefrod ftvfcurfhrohhfihhlvgemucfhrghsthforghilhdpqfgfvfdpuffrtefokffrpgfnqfgh necuuegrihhlohhuthemuceftddtnecupfhoucgurghtvgcufhhivghlugculdegledmne cujfgurheptggghffvufesrgdttdertddtjeenucfhrhhomheptfgvphhoshhithhorhih ucettghtihhvihhthicuufhumhhmrghrhicuuehothcuoeguohgpnhhothgprhgvphhlhi esmhhnohhtrdhnvghtqeenucggtffrrghtthgvrhhnpeekfedvudetjedvfeekheeiveeu gfefhfetteevgeffkefffeetffdvleehudeiteenucffohhmrghinhepghhithhhuhgsrd gtohhmnecuvehluhhsthgvrhfuihiivgeptdenucfrrghrrghmpehmrghilhhfrhhomhep ughopghnohhtpghrvghplhihsehmnhhothdrnhgvth
X-ME-Proxy: <xmx:y1bRYSuHv6_lPjpE7tLU7nn9gWB6SUia6ybqeUgeg65W5GSJhkMlNA> <xmx:y1bRYafz5AkFVKAjepSZSw1Hmd6rDM9sbSw1exCiOqQWI52sWE0WnA> <xmx:y1bRYW1kPtpxRjq-w-_sGRU6F9vz3ghtuJlDxmOxGteamZmywk3KSQ> <xmx:zFbRYbouxiY67TIl6Gxg-4BxwAHikVM7Mvq5spIM-5uoVv3cVtu9KQ>
Received: by mail.messagingengine.com (Postfix) with ESMTPA for <ppm@ietf.org>;  Sun, 2 Jan 2022 02:39:55 -0500 (EST)
Content-Type: multipart/alternative; boundary="===============6418912167448205287=="
MIME-Version: 1.0
From: Repository Activity Summary Bot <do_not_reply@mnot.net>
To: ppm@ietf.org
Message-Id: <20220102075555.4D6D03A1F72@ietfa.amsl.com>
Date: Sat,  1 Jan 2022 23:55:55 -0800 (PST)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/mAOPksrURKt6_xrEJXg4gsKApFA>
Subject: [Ppm] Weekly github digest (Privacy Preserving Measurement Activity Summary)
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 02 Jan 2022 07:56:01 -0000

--===============6418912167448205287==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"; format="flowed"




Events=20

Issues
------
* abetterinternet/ppm-specification (+0/-5/=F0=9F=92=AC15)
  4 issues received 15 new comments:
  - #173 addressing collusion (5 by cjpatton, ekr, npdoty, tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/173=20
  - #161 Rotation of shared secrets in PPM task parameters (4 by cjpatton, =
henrycg)
    https://github.com/abetterinternet/ppm-specification/issues/161=20
  - #139 Clarify encoding, endianness of multi-byte values in HPKE applicat=
ion info and AAD strings (5 by cjpatton, ekr, tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/139=20
  - #82 Allow batch "replay" as required by the PDA protocol (was "Inter-ba=
tch replay prevention") (1 by cjpatton)
    https://github.com/abetterinternet/ppm-specification/issues/82=20

  5 issues closed:
  - Cross-aggregator data consistency  https://github.com/abetterinternet/p=
pm-specification/issues/4 [duplicate]=20
  - Revisiting the security model https://github.com/abetterinternet/ppm-sp=
ecification/issues/86=20
  - The wrapper-core protocol interface https://github.com/abetterinternet/=
ppm-specification/issues/98=20
  - Clarify encoding, endianness of multi-byte values in HPKE application i=
nfo and AAD strings https://github.com/abetterinternet/ppm-specification/is=
sues/139=20
  - Allow batch "replay" as required by the PDA protocol (was "Inter-batch =
replay prevention") https://github.com/abetterinternet/ppm-specification/is=
sues/82=20



Pull requests
-------------
* abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC0)
  2 pull requests submitted:
  - nits (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/175=20
  - Have clients upload report shares to helpers (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/174=20

  1 pull requests merged:
  - nits
    https://github.com/abetterinternet/ppm-specification/pull/175=20


Repositories tracked by this digest:
-----------------------------------
* https://github.com/abetterinternet/ppm-specification

--===============6418912167448205287==
Content-Type: text/html; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html lang=3D"en">
<head>
<meta charset=3D"utf-8">
<title>Weekly github digest (Privacy Preserving Measurement Activity Summar=
y)</title>
<style>
body { font-family: Gotham, "Helvetica Neue", Helvetica, Arial, sans-serif;=
 font-size: 14px; }
h2 { margin-top: 3em; color: #A52A2A; font-style: italic; font-weight: norm=
al; }
h3 { margin-bottom:0; margin-top: 2em; font-size: 1.2em; }
h1+h2 { margin-top: 1em; }
a { color: #bb6219; text-decoration: none; }
li { margin-bottom: .35em; }
.repos { margin-bottom: 0; margin-top:0; line-height: 1.2; }
.new { color: red; }
.label { display: inline;
	padding: .2em .6em .3em;
	font-size: 75%;
	font-weight: 700;
	line-height: 1;
	color: #fff;
	text-align: center;
	white-space: nowrap;
	vertical-align: baseline;
	border-radius: .25em;
}
details {
	margin-top: 8em;
	}
summary {
	margin-bottom: 1em;
	cursor: pointer;
}
</style>
</head>

<body>
<h1>Sunday January 02, 2022</h1>

<p>Events </p>

<h2>Issues</h2>

<h3>abetterinternet/ppm-specification (+0/-5/=F0=9F=92=AC15)</h3>

  <p>4 issues received 15 new comments:</p>
  <ul>
  <li>#173 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/173">addressing collusion</a> (5 by cjpatton, ekr, npdoty, tgeoghega=
n) </li>
 =20
  <li>#161 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/161">Rotation of shared secrets in PPM task parameters</a> (4 by cjp=
atton, henrycg) </li>
 =20
  <li>#139 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/139">Clarify encoding, endianness of multi-byte values in HPKE appli=
cation info and AAD strings</a> (5 by cjpatton, ekr, tgeoghegan) </li>
 =20
  <li>#82 <a href=3D"https://github.com/abetterinternet/ppm-specification/i=
ssues/82">Allow batch &quot;replay&quot; as required by the PDA protocol (w=
as &quot;Inter-batch replay prevention&quot;)</a> (1 by cjpatton) </li>
  </ul>

  <p>5 issues closed:</p>
  <ul>
  <li>#4 <a href=3D"https://github.com/abetterinternet/ppm-specification/is=
sues/4">Cross-aggregator data consistency </a> <span class=3D"label" style=
=3D"background-color: #ffffff; color: #000000">duplicate</span> </li>
 =20
  <li>#86 <a href=3D"https://github.com/abetterinternet/ppm-specification/i=
ssues/86">Revisiting the security model</a> </li>
 =20
  <li>#98 <a href=3D"https://github.com/abetterinternet/ppm-specification/i=
ssues/98">The wrapper-core protocol interface</a> </li>
 =20
  <li>#139 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/139">Clarify encoding, endianness of multi-byte values in HPKE appli=
cation info and AAD strings</a> </li>
 =20
  <li>#82 <a href=3D"https://github.com/abetterinternet/ppm-specification/i=
ssues/82">Allow batch &quot;replay&quot; as required by the PDA protocol (w=
as &quot;Inter-batch replay prevention&quot;)</a> </li>
  </ul>



<h2>Pull requests</h2>
<h3>abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC0)</h3>
  <p class=3D"new">2 pull requests submitted:</p>
  <ul>
  <li>#175 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/175">nits</a> (by cjpatton) </li>
 =20
  <li>#174 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/174">Have clients upload report shares to helpers</a> (by cjpatton) </=
li>
  </ul>


  <p>1 pull requests merged:</p>
  <ul>
  <li>#175 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/175">nits</a> </li>
  </ul>


  <details>
    <summary>Repositories tracked by this digest:</summary>
<ul class=3D"repos">
  <li><a href=3D"https://github.com/abetterinternet/ppm-specification">http=
s://github.com/abetterinternet/ppm-specification</a></li>
</ul>
</details>
</body>
</html>

--===============6418912167448205287==--


From nobody Mon Jan  3 09:14:55 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 5261F3A0919 for <ppm@ietfa.amsl.com>; Mon,  3 Jan 2022 09:14:53 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -0.2
X-Spam-Level: 
X-Spam-Status: No, score=-0.2 tagged_above=-999 required=5 tests=[BAYES_40=-0.001, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id HhqE44CTpTWH for <ppm@ietfa.amsl.com>; Mon,  3 Jan 2022 09:14:49 -0800 (PST)
Received: from mail-qt1-x829.google.com (mail-qt1-x829.google.com [IPv6:2607:f8b0:4864:20::829]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 271713A0917 for <ppm@ietf.org>; Mon,  3 Jan 2022 09:14:49 -0800 (PST)
Received: by mail-qt1-x829.google.com with SMTP id f9so31048500qtk.4 for <ppm@ietf.org>; Mon, 03 Jan 2022 09:14:49 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to; bh=FSG6Yw9QmhWr63UWTruRgQWY5AJThUe83xoHy26OSyQ=; b=TAKLHyzINQbbpFZG3vcopW59VfqlDSWBHRWic85hIEfrq0Wzz7pFIZxyDeGK35x96M 66bmJK39682u37MtkxSt5i846AEE4+GiIdP0YVQPDQ11CzkLUsuBiWJW3XV0nAFg04Vu IvIBsElCg5IQgQCPHTwabddZUkHtikA0qHLlY=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to; bh=FSG6Yw9QmhWr63UWTruRgQWY5AJThUe83xoHy26OSyQ=; b=t/AnBxs9HDtKiUO16E/1aHhRFDXECNKX9Ol5sLCZofBCFSrJ9lspds77P+Tfe6sLEj rkNo49DWrgBMAHRjq7UNs6KC9vM+AMsQIRoEhmn4OsnpggaqhFpwcJhdCAEtRX52rySX roshrHpqHrzMuSenDeiOHAT+CuZXn9NZ37BcvK009eR4AYSL89SwxzKNns2wgWbu4iU3 A2VcUZDOjjt859kKhm9of/u4278mRqACzswVEZ4Upo9v/z3qdLSiSppZ4cu8f59Ll0OE taDWYL4I4/ISx6KwfoGXnVbwEYad0cKoOkXdLqHEWhjWNSdt23EnY/VBo9cJXgoBrYil ZJzA==
X-Gm-Message-State: AOAM532UjWZu5HXbHS2L+JPKPa8Fo0NPSZKniS3pjeLFubjgpW4n/jNZ 4ByZOZbbbhGpf2rQSH1q7oEFJ9maEYxXUCfcy98IWjFrFJN2+w==
X-Google-Smtp-Source: ABdhPJy1yWcDz6VQHhF8NKAJfgSCoXerYx3UuF6slZWtE06YD+CLwmOkw3/dMGZd46eySRDI1IDdfaInv9QKMtldOXA=
X-Received: by 2002:a05:622a:389:: with SMTP id j9mr41964394qtx.504.1641230087155;  Mon, 03 Jan 2022 09:14:47 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi20oarjT6jBbtFtvQUMy5uRZTK6Z2YgBSE+wT33xQbkb+A@mail.gmail.com>
In-Reply-To: <CAG2Zi20oarjT6jBbtFtvQUMy5uRZTK6Z2YgBSE+wT33xQbkb+A@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Mon, 3 Jan 2022 09:14:36 -0800
Message-ID: <CAG2Zi21N3fHZpvXadLbt3rhC-5LzLVoPRZdj01ONJCj-4YsfdA@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000057110e05d4b0a8d3"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/CfSDiW19sPSKaIhShdB55jzvRcY>
Subject: Re: [Ppm] Revisit the operational requirements for the Helper?
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 03 Jan 2022 17:14:53 -0000

--00000000000057110e05d4b0a8d3
Content-Type: text/plain; charset="UTF-8"

Happy new year friends! The PR with the suggested change to anti-replay is
up to date and ready for another round of review:
https://github.com/abetterinternet/ppm-specification/pull/169

It would be great if, by the end of the week, we can get a sense of whether
folks support (some version of) this change or not

Cheers,
Chris P.

On Thu, Dec 2, 2021 at 4:04 PM Christopher Patton <cpatton@cloudflare.com>
wrote:

> Hi everyone,
>
> I've begun an implementation of the PPM spec with `prio3` [1]. One of the
> design goals of the sepc was to make it as easy as possible for a Helper to
> participate in the protocol, even at the expense of higher resource
> requirements for the Leader. Based on my experience so far, I think we may
> have over-indexed on this goal a bit. I propose that we revisit this and
> consider balancing the work between the Aggregators more evenly. While at
> it, we may be able to resolve [2].
>
> The spec requires that reports be processed in strictly ascending order,
> where the order is defined in [3]. The ordering is based on a timestamp and
> random number generated by the client and included in the report. I'll
> refer to thus (time, rand) pair as the nonce. Ordering reports based on the
> nonce allows the Helper to prevent replay attacks* without needing to store
> the nonces. In fact, the only long-term storage it needs for anti-replay is
> the most recently processed nonce. The downside is that the Leader has to
> sort reports as they arrive, which makes horizontally scaling the Leader
> more challenging.
>
> A simpler alternative that we had discussed previously is just having each
> of the Aggregators remember the nonces of the reports they have processed.
> This removes the Leader's coordination problem, at the cost of additional
> storage requirements. In fact, both Aggregators need to store the set of
> processed nonces for the duration of a task. This cost can be mitigated by
> refusing to process reports that arrive, say, a day or two late. (Of
> course, this would need to be a parameter that the Aggregators agree on in
> advance.)
>
> Let's call these schemes option #1 and option #2 respectively. Previously
> I was in favor of option #1, but after getting my hands a little dirty and
> conferring with some of my colleagues, I've come to the conclusion that the
> costs of #1 outweigh the costs of #2. I'm curious if others have thought
> about this and whether they are strongly in favor of either option.
>
> Now, if we take option #2, we might also consider revisiting the question
> raised in [2]. The spec currently requires the Clients to upload their
> encrypted input shares to the Leader, who then piggy-backs the Helper's
> input shares on its aggregate requests. Egress fees from cloud providers is
> typically higher than ingress, which makes the status quo more expensive to
> deploy than if the Clients uploaded their encrypted input shares directly
> to each Aggregator.
>
> What might an alternative look like? The client would generate a nonce as
> usual and upload the nonce and an encrypted input share to each of the
> Aggregators. The aggregation flow would largely be the same: The Leader
> sends an Aggregate Request to the Helper, which contains the nonce of the
> report it wants to aggregate as usual. The difference now is that the
> Helper may not have yet received the report before the Leader's request. In
> this case the Leader would just try again later.
>
> Another advantage of this alternative is that it prevents a fairly weak
> Sybil attack, whereby the Leader makes up `n-1` bogus reports that are
> processed into the batch.**
>
>
> Thanks for your consideration,
> Chris P.
>
>
> * Anti-replay is required to prevent privacy violations. See [3].
> ** The Leader can still mount a slightly stronger attack: Just upload the
> `n-1` encrypted input shares to the Helper, impersonating the Clients.
> Though this can be mitigated slightly by rate-limiting requests from the
> Leader's IP address.
>
> [1]
> https://datatracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#section-6
> [2] https://github.com/abetterinternet/ppm-specification/issues/130
> [3]
> https://datatracker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2
>
>

--00000000000057110e05d4b0a8d3
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Happy new year friends! The PR with the suggested change t=
o anti-replay is up to date and ready for another round of review:<br><div>=
<a href=3D"https://github.com/abetterinternet/ppm-specification/pull/169">h=
ttps://github.com/abetterinternet/ppm-specification/pull/169</a></div><div>=
<br></div><div> It would be great if, by the end of the week, we can get a =
sense of whether folks support (some version of) this change or not</div><d=
iv><br></div><div>Cheers,</div><div>Chris P.<br> </div></div><br><div class=
=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Dec 2, 2021 =
at 4:04 PM Christopher Patton &lt;<a href=3D"mailto:cpatton@cloudflare.com"=
>cpatton@cloudflare.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_=
quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,=
204);padding-left:1ex"><div dir=3D"ltr"><div>Hi everyone,</div><div><br></d=
iv><div>I&#39;ve begun an implementation of the PPM spec with `prio3` [1]. =
One of the design goals of the sepc was to make it as easy as possible for =
a Helper to participate in the protocol, even at the expense of higher reso=
urce requirements for the Leader. Based on my experience so far, I think we=
 may have over-indexed on this goal a bit. I propose that we revisit this a=
nd consider balancing the work between the Aggregators more evenly. While a=
t it, we may be able to resolve [2].</div><div><br></div><div>The spec requ=
ires that reports be processed in strictly ascending order, where the order=
 is defined in [3]. The ordering is based on a timestamp and random number =
generated by the client and included in the report. I&#39;ll refer to thus =
(time, rand) pair as the nonce. Ordering reports based on the nonce allows =
the Helper to prevent replay attacks* without needing to store the nonces. =
In fact, the only long-term storage it needs for anti-replay is the most re=
cently processed nonce. The downside is that the Leader has to sort reports=
 as they arrive, which makes horizontally scaling the Leader more challengi=
ng.</div><div><br></div><div>A simpler alternative that we had discussed pr=
eviously is just having each of the Aggregators remember the nonces of the =
reports they have processed. This removes the Leader&#39;s coordination pro=
blem, at the cost of additional storage requirements. In fact, both Aggrega=
tors need to store the set of processed nonces for the duration of a task. =
This cost can be mitigated by refusing to process reports that arrive, say,=
 a day or two late. (Of course, this would need to be a parameter that the =
Aggregators agree on in advance.)<br></div><div><br></div><div>Let&#39;s ca=
ll these schemes option #1 and option #2 respectively. Previously I was in =
favor of option #1, but after getting my hands a little dirty and conferrin=
g with some of my colleagues, I&#39;ve come to the conclusion that the cost=
s of #1 outweigh the costs of #2. I&#39;m curious if others have thought ab=
out this and whether they are strongly in favor of either option.</div><div=
><br></div><div>Now, if we take option #2, we might also consider revisitin=
g the question raised in [2]. The spec currently requires the Clients to up=
load their encrypted input shares to the Leader, who then piggy-backs the H=
elper&#39;s input shares on its aggregate requests. Egress fees from cloud =
providers is typically higher than ingress, which makes the status quo more=
 expensive to deploy than if the Clients uploaded their encrypted input sha=
res directly to each Aggregator.</div><div><br></div><div>What might an alt=
ernative look like? The client would generate a nonce as usual and upload t=
he nonce and an encrypted input share to each of the Aggregators. The aggre=
gation flow would largely be the same: The Leader sends an Aggregate Reques=
t to the Helper, which contains the nonce of the report it wants to aggrega=
te as usual. The difference now is that the Helper may not have yet receive=
d the report before the Leader&#39;s request. In this case the Leader would=
 just try again later.<br></div><div><br></div><div>Another advantage of th=
is alternative is that it prevents a fairly weak Sybil attack, whereby the =
Leader makes up `n-1` bogus reports that are processed into the batch.**</d=
iv><div><br></div><div><br></div><div>Thanks for your consideration,</div><=
div>Chris P.<br></div><div></div><div><br></div><div><br></div><div>* Anti-=
replay is required to prevent privacy violations. See [3].</div><div>** The=
 Leader can still mount a slightly stronger attack: Just upload the=20
`n-1` encrypted input shares to the Helper, impersonating the Clients. Thou=
gh this can be mitigated slightly by rate-limiting requests from the Leader=
&#39;s IP address.<br></div><div><br></div><div>[1] <a href=3D"https://data=
tracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#section-6" target=3D"_b=
lank">https://datatracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#secti=
on-6</a></div><div>[2] <a href=3D"https://github.com/abetterinternet/ppm-sp=
ecification/issues/130" target=3D"_blank">https://github.com/abetterinterne=
t/ppm-specification/issues/130</a></div><div>[3] <a href=3D"https://datatra=
cker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2" target=3D"_blank"=
>https://datatracker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2</a=
></div><div><br></div></div>
</blockquote></div>

--00000000000057110e05d4b0a8d3--


From nobody Tue Jan  4 11:09:54 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id AFE923A09D2 for <ppm@ietfa.amsl.com>; Tue,  4 Jan 2022 11:09:52 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id fX8HP57TEyP1 for <ppm@ietfa.amsl.com>; Tue,  4 Jan 2022 11:09:48 -0800 (PST)
Received: from mail-qt1-x82f.google.com (mail-qt1-x82f.google.com [IPv6:2607:f8b0:4864:20::82f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 6BE2D3A0ACB for <ppm@ietf.org>; Tue,  4 Jan 2022 11:09:48 -0800 (PST)
Received: by mail-qt1-x82f.google.com with SMTP id j17so35151922qtx.2 for <ppm@ietf.org>; Tue, 04 Jan 2022 11:09:48 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:from:date:message-id:subject:to; bh=wtTerietG/yIZXc5/2KwtsyqTXC2wgMwVQ5GQOrYTOM=; b=NGkE2yCVlajkAZFeIGaBbrsIEiD1USAc3/cq9VqyswJ35AQ3jTSOPasXw7RLFdppYP +O9bY455JfpJsZBS012Pd4f3gTdXJuJ+XJnWRKxpvgKclqmOrid9RaCjQIFJ0F9OnJmd gjDVFpewthqVAJTMUHsMEWyQrg9Bc2oH7iP00=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:from:date:message-id:subject:to; bh=wtTerietG/yIZXc5/2KwtsyqTXC2wgMwVQ5GQOrYTOM=; b=0e5fBgxP5Dj6lyCL/4IIFCMiU/ZBcp1K2l8JPR6P97xMPTHxSYz9XRI8+CYBHQWw0Z 0GzUnBzGGEebu4z06Hz3WhtKXOTGatNEL1RdRkn+p8SSL/0iS2q0j3GuWAnyaVkn29Cu BKPfM+FGclkww1Un1T9puzqdnP6JkRu3LordaIouCAsKDUKVydai63jkIF5toej/933a 68zJY+PI6cR4eTjGl9IdACEx2HREYQe28SwTaXbGOzHununVnPl/VlJ/9pyjGFkshg1k wuz3eBdT5cEG3cukwqkqHGpJWP98XvCrlorbAqTqspPC/KXEkHphK97YuIeynHTmIg0g C89Q==
X-Gm-Message-State: AOAM532RHntZ/OYzyGXgG1XEayzskVLTUJzFME6hCnfAl6+d0rsLUF9V Ejk+WI0vxsiu5yUQCFM8NCoAHXmXlI3yhN4ihjWH2+cNAX/VeA==
X-Google-Smtp-Source: ABdhPJzRoBFbR8QJVyY7pwtKxIFGcxfq9bKvW/Qwvpw0S4aPgm0SbgnJvtx2QH/R+jnecIXQNtPneV08cSJDfD4gQ5s=
X-Received: by 2002:ac8:5e4d:: with SMTP id i13mr45778770qtx.452.1641323385493;  Tue, 04 Jan 2022 11:09:45 -0800 (PST)
MIME-Version: 1.0
From: Christopher Patton <cpatton@cloudflare.com>
Date: Tue, 4 Jan 2022 11:09:34 -0800
Message-ID: <CAG2Zi20Wg1wDfbDW9tF2AkLfQ9BkQy+50e9kV3R2-0L69D=7Tw@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000005ab8d905d4c661fa"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/7S8NaCeD37sM_ArusaNlBIiamO8>
Subject: [Ppm] "Heavy Hitters" is now "Poplar"
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 04 Jan 2022 19:09:53 -0000

--0000000000005ab8d905d4c661fa
Content-Type: text/plain; charset="UTF-8"

As noted here, we now have a proper name for the privacy-preserving heavy
hitters protocol:
https://github.com/abetterinternet/ppm-specification/issues/176

I will create a PR to update the name. Unless there's an objection, I'd
like to merge it by the end of the week.

Best,
Chris P.

--0000000000005ab8d905d4c661fa
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>As noted here, we now have a proper name for the priv=
acy-preserving heavy hitters protocol: <a href=3D"https://github.com/abette=
rinternet/ppm-specification/issues/176">https://github.com/abetterinternet/=
ppm-specification/issues/176</a></div><div><br></div><div>I will create a P=
R to update the name. Unless there&#39;s an objection, I&#39;d like to merg=
e it by the end of the week.</div><div><br></div>Best,<div>Chris P.<br></di=
v></div>

--0000000000005ab8d905d4c661fa--


From nobody Wed Jan  5 13:19:33 2022
Return-Path: <rdd@cert.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id A4DBA3A0C6E for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:19:31 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.9
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=seicmu.onmicrosoft.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id m-YdAtcUqgys for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:19:27 -0800 (PST)
Received: from USG02-BN3-obe.outbound.protection.office365.us (mail-bn3usg02on0091.outbound.protection.office365.us [23.103.208.91]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 525403A0C6B for <ppm@ietf.org>; Wed,  5 Jan 2022 13:19:26 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector5401; d=microsoft.com; cv=none; b=pxw1sS8HfHLfGD9Naa1dF1tCXywZwT+8htEYcAKEawVZGgeEjTn8mLC3/tc3c01zI+up4k4eYfz9S6WzWNyeOwLuzQ5rmvBcynSWa5nigQE7WBTcYy9A6uHAWOA9weNeNcopyKady+mEyZOYpckuL+eplE1XDVkbMIs4LQAVrlkKI+buX8LdsaMBpjeNlAreQj5UTU0KE26CwBwvvcpCukB/pEjzs61PFDDi7D5Pc6KyEW3ZTtVWUckw8wCUBS2gWcsB3nQhWhK1iVgTYDL0KOsxqbwMNH044FG/FE/3C4o5MzMqOKCrT5zp0BcxgBzX96MIoxbHNSBk3g8RA6tXQg==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector5401; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=qrYqYhb21kafQqXZ3emCF+7B7eMoXs4Wpo16q7IHaEI=; b=UywDv7CSxIbSH/POCy/0Af2h+ALT3qcCS4JZcbJFf+JZe856g/47QBRNqOok+9N7d9OYfWdYw2qXBpxsDVm/tBaZwefdvC3UMCs9meU1gQahGtMitzEua3WgaZ6r2sLYa4zA7WHO47esBPIg3vdEaezw5P7vQ4/vL2cSnN6lDl1iKt5ww8sabWu8m2AxDf3dbL7l2ztoFNVZr5y6e1jJR7nEE4zL4Sz8tkWUwvZb5/wIyIqPRFOigE8F+/JhK8n5NjjKxWqsG2b45UFhHrzVZtpygpeVuaAZ1dwB65C/qGKsLTDvLOSuwXLgQWyZPnd+exGUfmM4dwDT4bNNmJ9V/w==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cert.org; dmarc=pass action=none header.from=cert.org; dkim=pass header.d=cert.org; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=seicmu.onmicrosoft.com; s=selector1-seicmu-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=qrYqYhb21kafQqXZ3emCF+7B7eMoXs4Wpo16q7IHaEI=; b=csPXg1XrPaTfnIUeq9TA646H2FkRXUqeAqdchGFJGKPUcU1uBah0xpWhMiKGIk2zwkE/aDwKnTVlAwCdwwLftXI4O+ZUqYMi2NxOFav7yKw7wqwT4Wll5andnuqHkDVn208awOTKs9MbyBxB8yy/yTjdNCzhGwUN7ZkI8EIyHyQ=
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::11) by BN2P110MB1223.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:17c::15) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4844.14; Wed, 5 Jan 2022 21:19:23 +0000
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::f0be:6d5:6544:cce0]) by BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::f0be:6d5:6544:cce0%4]) with mapi id 15.20.4844.016; Wed, 5 Jan 2022 21:19:23 +0000
From: Roman Danyliw <rdd@cert.org>
To: "ppm@ietf.org" <ppm@ietf.org>
Thread-Topic: Call for consensus on updated charter
Thread-Index: AdgCd8GCV7y9nJtBRKqSR9LxS59tQQ==
Date: Wed, 5 Jan 2022 21:19:23 +0000
Message-ID: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cert.org;
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 0672dc50-35e6-46b1-0e1f-08d9d0910bdb
x-ms-traffictypediagnostic: BN2P110MB1223:
x-microsoft-antispam-prvs: <BN2P110MB12239E453CBC5A7280A3500CDC4B9@BN2P110MB1223.NAMP110.PROD.OUTLOOK.COM>
x-ms-oob-tlc-oobclassifiers: OLM:10000;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM; PTR:; CAT:NONE;  SFS:(366004)(52536014)(82960400001)(498600001)(8936002)(86362001)(55016003)(71200400001)(8676002)(186003)(66946007)(38070700005)(83380400001)(6506007)(2906002)(76116006)(9686003)(6916009)(66446008)(7696005)(966005)(66476007)(5660300002)(38100700002)(64756008)(33656002)(26005)(122000001)(66556008); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: F/HdBBl5JjtExzjBopcOR6y5f7jGuu8/wYZS0Aq/ITZfZZUrgLLUz6ynPPUs5A1e9sRF90ULtGPIjYUYmHjZUT6f8BtK93Gnj91eXbNaOqZQe1a0OIcoa5rIRWXBdP/Qm7kTG4KMM9j2Cig6sG/Pig==
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: cert.org
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM
X-MS-Exchange-CrossTenant-Network-Message-Id: 0672dc50-35e6-46b1-0e1f-08d9d0910bdb
X-MS-Exchange-CrossTenant-originalarrivaltime: 05 Jan 2022 21:19:23.6460 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 95a9dce2-04f2-4043-995d-1ec3861911c6
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN2P110MB1223
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/-H3AEHSiQiQ2voW4BtxBLLvci_s>
Subject: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 05 Jan 2022 21:19:32 -0000

Hi!

Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF 11=
2.  The meeting notes [1] captured four major themes of feedback based on t=
he presented charter [2] (version 00-00).  Specifically:

=3D=3D[ snip from meeting notes ]=3D=3D
(a) Generalize the text to make it clear that flexibility for alternatives =
and additions to approach is possible (i.e., more than Prio, heavy hitters)
(b) Refine the definition of aggregations so as to not restrict alternative=
s
(c) Add a work item(s) to document the abuse cases and associated mitigatio=
ns from the perspective of all participants in the architecture
(d) Revision of the WG name to convey a more narrow scope
=3D=3D[ snip ]=3D=3D =20

In response to this feedback, the proposed charter text has been revised at=
 [3] (version 00-01).  Also, see [4] for a diff of this revision against th=
e original text presented at IETF 112.  For convenience, this charter text =
is loaded into the datatracker using the BOF name (PRIV). As there wasn't c=
onsensus on "PRIV" as the WG name per feedback (d) or the now proposed alte=
rnative of "PPM", it didn't make sense to ask the Secretariat to create a n=
ew entry in the datatracker yet.  If consensus is found, a new datatracker =
entry will be made prior to sending the charter to the IESG.

This email is intended to start a thread to gauge interest in forming a WG =
around this new, refined charter language [3].  Please respond to the list:

(1) Do you support the charter text [3]? Or do you have objections or block=
ing concerns (please describe what they might be)?

If you do support the charter text:
(2) Are you willing to author or participate in the developed of the WG dra=
fts?
(3) Are you willing to review the WG drafts?
(4) Are you interested in implementing the WG drafts?

This call for consensus will end on Thursday, January 20.

Regards,
Roman=20

[1] https://datatracker.ietf.org/doc/minutes-112-priv/
[2] charter-ietf-priv-00-00 - https://datatracker.ietf.org/doc/charter-ietf=
-priv/00-00/
[3] charter-ietf-priv-00-01 - https://datatracker.ietf.org/doc/charter-ietf=
-priv/00-01/
[4] diff of 00-00 vs. 00-01 -- https://www.ietf.org/rfcdiff?url1=3Dhttps%3A=
%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00=
.txt&url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2F=
withmilestones-00-01.txt
[5] https://datatracker.ietf.org/wg/priv/about/


From nobody Wed Jan  5 13:39:43 2022
Return-Path: <ekr@rtfm.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 995FB3A0D26 for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:39:41 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.896
X-Spam-Level: 
X-Spam-Status: No, score=-1.896 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_NONE=0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=rtfm-com.20210112.gappssmtp.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ofWlR4FgvGII for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:39:36 -0800 (PST)
Received: from mail-il1-x131.google.com (mail-il1-x131.google.com [IPv6:2607:f8b0:4864:20::131]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id AE1B83A0D10 for <ppm@ietf.org>; Wed,  5 Jan 2022 13:39:36 -0800 (PST)
Received: by mail-il1-x131.google.com with SMTP id d14so603757ila.1 for <ppm@ietf.org>; Wed, 05 Jan 2022 13:39:36 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=rtfm-com.20210112.gappssmtp.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=yFjLg8IXKDMeNcu1xKg6cetu1aAveBVOyKj7cLBma7k=; b=7RCQppBbATmP7HotBIynZ0UVjBKJCB9/6WshVvShh12QH+ue66pF49zqo3n37ZRcNY JshNPygW9IRk/aKdF2bwEVc9ApMlpYbZ3gOh7GpBNQqSNptMWnVuY2r/0CACyHLA+YcN dxeiN0+n0mQpxCY/pmp/ZfZRA9NYRYItdPOKrcjLR2bHcRBzkvP1aDb9LdedRHLYDMPK rrnJkrkSwmql9r7zxPm0NRjcgbILlXilfIlz1gIrdD9Jrjyi4YKjbQUSlMzHXR36HT2r xtn2aSmibouJuszIvlbcB8a4xD+xJlO7bfqtuiV2y+qggzwGb02/2+t+rbSpNvy3ZG5/ iWsA==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=yFjLg8IXKDMeNcu1xKg6cetu1aAveBVOyKj7cLBma7k=; b=0P9ohUD9AvuahfT2hcMc3s2hDcjSD4IWDOmiKaH60TmcMmQONgBwFqG8gHNhXZJ/So k4nEvc15KeD632mt1OLmfMu/NxV2eg6fQgYhlTkFCfrGdqHhMro+2pBQ7rWt5u6Zub4v 9CmxcjxCpUwYmNc7wDanJT8LkdHJw8xdgCir+aWymncAFr3t73AprDq1Uolhr8zoaOhd cxfnxzSPJktYXECIg26S1GZF99kSc2jNlimYbl4CEFuJ/XY6IfcCjBsnEAGi+5rFQWKb PfDYsukHdR+JKVsZqaid+j80AKRscNtWvnUTfkBUGMwNKcHS2QpntvadqIlt/kPGiSsV tMqQ==
X-Gm-Message-State: AOAM530J3W2hC/4IvEHKDve4YsHqO6uwaNFVuGRlFwOyhRMLBQwysuCo +Mb1vXbNo5OLueCXBxPBfLzJXOlEw3wqgWNBrkBNP26xFVQ=
X-Google-Smtp-Source: ABdhPJzJWeucf3xsgdPMTx5Vrz5kbARjy6G8El1fp+QYJUfSOTvXD+anV9WqGuwEbIpreSbYtk496Ad16j+o1zyFpms=
X-Received: by 2002:a05:6e02:14d1:: with SMTP id o17mr26803818ilk.276.1641418774488;  Wed, 05 Jan 2022 13:39:34 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Eric Rescorla <ekr@rtfm.com>
Date: Wed, 5 Jan 2022 13:38:58 -0800
Message-ID: <CABcZeBPNAJF5L1vaxo8t2vP3K897c3OLC9udD7opqdG+4f3-2g@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="000000000000fb413905d4dc9673"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/oqjqIN6dk7drVOWev0OlNffcTjo>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 05 Jan 2022 21:39:42 -0000

--000000000000fb413905d4dc9673
Content-Type: text/plain; charset="UTF-8"

On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw <rdd@cert.org> wrote:

> Hi!
>
> Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF
> 112.  The meeting notes [1] captured four major themes of feedback based on
> the presented charter [2] (version 00-00).  Specifically:
>
> ==[ snip from meeting notes ]==
> (a) Generalize the text to make it clear that flexibility for alternatives
> and additions to approach is possible (i.e., more than Prio, heavy hitters)
> (b) Refine the definition of aggregations so as to not restrict
> alternatives
> (c) Add a work item(s) to document the abuse cases and associated
> mitigations from the perspective of all participants in the architecture
> (d) Revision of the WG name to convey a more narrow scope
> ==[ snip ]==
>
> In response to this feedback, the proposed charter text has been revised
> at [3] (version 00-01).  Also, see [4] for a diff of this revision against
> the original text presented at IETF 112.  For convenience, this charter
> text is loaded into the datatracker using the BOF name (PRIV). As there
> wasn't consensus on "PRIV" as the WG name per feedback (d) or the now
> proposed alternative of "PPM", it didn't make sense to ask the Secretariat
> to create a new entry in the datatracker yet.  If consensus is found, a new
> datatracker entry will be made prior to sending the charter to the IESG.
>
> This email is intended to start a thread to gauge interest in forming a WG
> around this new, refined charter language [3].  Please respond to the list:
>
> (1) Do you support the charter text [3]? Or do you have objections or
> blocking concerns (please describe what they might be)?
>
> If you do support the charter text:
>

Yes.



> (2) Are you willing to author or participate in the developed of the WG
> drafts?
>

Yes


> (3) Are you willing to review the WG drafts?
>

Yes


> (4) Are you interested in implementing the WG drafts?
>

Yes

>
> This call for consensus will end on Thursday, January 20.
>
> Regards,
> Roman
>
> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> [2] charter-ietf-priv-00-00 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-00/
> [3] charter-ietf-priv-00-01 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-01/
> [4] diff of 00-00 vs. 00-01 --
> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt
> [5] https://datatracker.ietf.org/wg/priv/about/
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000fb413905d4dc9673
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 5, 2022 at 1:19 PM Roman =
Danyliw &lt;<a href=3D"mailto:rdd@cert.org">rdd@cert.org</a>&gt; wrote:<br>=
</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;b=
order-left:1px solid rgb(204,204,204);padding-left:1ex">Hi!<br>
<br>
Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF 11=
2.=C2=A0 The meeting notes [1] captured four major themes of feedback based=
 on the presented charter [2] (version 00-00).=C2=A0 Specifically:<br>
<br>
=3D=3D[ snip from meeting notes ]=3D=3D<br>
(a) Generalize the text to make it clear that flexibility for alternatives =
and additions to approach is possible (i.e., more than Prio, heavy hitters)=
<br>
(b) Refine the definition of aggregations so as to not restrict alternative=
s<br>
(c) Add a work item(s) to document the abuse cases and associated mitigatio=
ns from the perspective of all participants in the architecture<br>
(d) Revision of the WG name to convey a more narrow scope<br>
=3D=3D[ snip ]=3D=3D=C2=A0 <br>
<br>
In response to this feedback, the proposed charter text has been revised at=
 [3] (version 00-01).=C2=A0 Also, see [4] for a diff of this revision again=
st the original text presented at IETF 112.=C2=A0 For convenience, this cha=
rter text is loaded into the datatracker using the BOF name (PRIV). As ther=
e wasn&#39;t consensus on &quot;PRIV&quot; as the WG name per feedback (d) =
or the now proposed alternative of &quot;PPM&quot;, it didn&#39;t make sens=
e to ask the Secretariat to create a new entry in the datatracker yet.=C2=
=A0 If consensus is found, a new datatracker entry will be made prior to se=
nding the charter to the IESG.<br>
<br>
This email is intended to start a thread to gauge interest in forming a WG =
around this new, refined charter language [3].=C2=A0 Please respond to the =
list:<br>
<br>
(1) Do you support the charter text [3]? Or do you have objections or block=
ing concerns (please describe what they might be)?<br>
<br>
If you do support the charter text:<br></blockquote><div><br></div><div>Yes=
.</div><div><br></div><div>=C2=A0<br></div><blockquote class=3D"gmail_quote=
" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);=
padding-left:1ex">
(2) Are you willing to author or participate in the developed of the WG dra=
fts?<br></blockquote><div><br></div><div>Yes</div><div>=C2=A0<br></div><blo=
ckquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex">
(3) Are you willing to review the WG drafts?<br></blockquote><div><br></div=
><div>Yes</div><div>=C2=A0<br></div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex">
(4) Are you interested in implementing the WG drafts?<br></blockquote><div>=
<br></div><div>Yes=C2=A0 <br></div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex">
<br>
This call for consensus will end on Thursday, January 20.<br>
<br>
Regards,<br>
Roman <br>
<br>
[1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=3D"n=
oreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-112-p=
riv/</a><br>
[2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-00/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-00/</a><br>
[3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-01/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-01/</a><br>
[4] diff of 00-00 vs. 00-01 -- <a href=3D"https://www.ietf.org/rfcdiff?url1=
=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmiles=
tones-00-00.txt&amp;url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fchart=
er-ietf-priv%2Fwithmilestones-00-01.txt" rel=3D"noreferrer" target=3D"_blan=
k">https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2=
Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&amp;url2=3Dhttps%3A%2F%=
2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt=
</a><br>
[5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"noreferr=
er" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div>

--000000000000fb413905d4dc9673--


From nobody Wed Jan  5 13:45:53 2022
Return-Path: <stephen.farrell@cs.tcd.ie>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 91E1A3A0FC6 for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:45:51 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.715
X-Spam-Level: 
X-Spam-Status: No, score=-2.715 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, NICE_REPLY_A=-0.714, RCVD_IN_MSPIKE_H2=-0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=cs.tcd.ie
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ukyn1hE9QvU3 for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 13:45:46 -0800 (PST)
Received: from EUR05-DB8-obe.outbound.protection.outlook.com (mail-db8eur05on2109.outbound.protection.outlook.com [40.107.20.109]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 774643A0FC4 for <ppm@ietf.org>; Wed,  5 Jan 2022 13:45:46 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=WeuPXxSbbS9NvVkbeevCcvD8xbBNBqvC8uCXrECD7LBC4dzR5Q6d4Pi3Y2oKB5mDvNBvEvCgi5J5Zpm0mic1wp4VoSOo5cxRATGtFsmQjyoi37K58KNfS6whWiGL+HTRq7hNjrZdWksa2LS6pZsG3H+1LQWj8SqVJcSi1ZEfgaSFuYepJrhGiwjz0EhBAiAdQjmJ1YIbKMdAiyRkdV8Rry2YP1nUx6qGyZ9PFwqsuhIgZBXOfIun/ROklbqPFeb/krVjYTvRZBh7yK5uUfGSpFZdhiG2D9nAEwYvQTG22e/zppfrFOuAbCBP5ErUVdUCwY083yauXC70DQ5WAvRnmg==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=N6MRekPetYk6bIxUghDGdeDiJD/FJjK5HgJIjTVq5r0=; b=E8vhlCV6FMYVOYy5NWTXgrgFNAxpfCGunDT7RkGh4nX71VkvmZnIxoafFM3LanO9wsiEvOZ5gJpkkjjw/pHJPU8RI2PYaw5GGwoMk8NAQovK9h33QRDLFztITcENi/ZXBHvALZlKWeNvb50T4g9Ljvl8XXfs4r5jWINcxQhZjCWlUntaXf5z5bla9N5qwyRbBgLBjbFaeTi1cmZvEi8vyAKdh99szpMp521I3DWHTWIKAq+q54L9Z8qrQHmi5GOioFMYlTEi7WETxVEKp30pBF0WYYmIhWdnZ4wAETCvHoFu4TDUhgEz81/TOPu+KMOvW2zQ2HHni72iklOiJ5IZxQ==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cs.tcd.ie; dmarc=pass action=none header.from=cs.tcd.ie; dkim=pass header.d=cs.tcd.ie; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cs.tcd.ie; s=selector1; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=N6MRekPetYk6bIxUghDGdeDiJD/FJjK5HgJIjTVq5r0=; b=jsSFxrL3qc7/ZpBmsZDSeK1+voKzxxFfAFrr6UpUGQBwG5ei+r5w+sgeMvgN6XGo4dCynWEe1xXl2hUQ4Wxoe5JL1137DWjdoDpQz2hIAGdpixYvCvcnVGjlQXhqQm1tGOrTpMDFtbYJK2XgvNhL/k12URO5kQei2xGVRWOtnM+3WCrLQzsi4AqcAoAQHIykzN94ut6C6Eue8YqFJojud8jM95YPhhAiP1SUF1/Ii22LnLRlczdj51Fzq7xeSQ7IEsuZotAKhmVrH52v5EKX4PW20grl0R0IWMLplQTBg1pdhfBaTTXfP+634v9Ta2QJBMiFMZ9fgg9ZlcoU2uQg/g==
Authentication-Results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cs.tcd.ie;
Received: from DB7PR02MB5113.eurprd02.prod.outlook.com (2603:10a6:10:77::15) by DB9PR02MB6652.eurprd02.prod.outlook.com (2603:10a6:10:21c::22) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4844.14; Wed, 5 Jan 2022 21:45:42 +0000
Received: from DB7PR02MB5113.eurprd02.prod.outlook.com ([fe80::a85c:e144:6533:99ea]) by DB7PR02MB5113.eurprd02.prod.outlook.com ([fe80::a85c:e144:6533:99ea%4]) with mapi id 15.20.4844.016; Wed, 5 Jan 2022 21:45:42 +0000
Message-ID: <9d699c4f-3b00-ab9a-f7bd-bfac507edc75@cs.tcd.ie>
Date: Wed, 5 Jan 2022 21:45:40 +0000
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Thunderbird/91.3.1
Content-Language: en-US
To: Roman Danyliw <rdd@cert.org>, "ppm@ietf.org" <ppm@ietf.org>
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Stephen Farrell <stephen.farrell@cs.tcd.ie>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="------------6iCS5SguaFTCuhCDKrELvYm0"
X-ClientProxiedBy: DB6PR07CA0074.eurprd07.prod.outlook.com (2603:10a6:6:2b::12) To DB7PR02MB5113.eurprd02.prod.outlook.com (2603:10a6:10:77::15)
MIME-Version: 1.0
X-MS-Exchange-MessageSentRepresentingType: 1
X-MS-PublicTrafficType: Email
X-MS-Office365-Filtering-Correlation-Id: 235e45e2-8283-48e8-070e-08d9d094b890
X-MS-TrafficTypeDiagnostic: DB9PR02MB6652:EE_
X-MS-Exchange-SharedMailbox-RoutingAgent-Processed: True
X-Microsoft-Antispam-PRVS: <DB9PR02MB665206E7CF58D91BEA7B324BA84B9@DB9PR02MB6652.eurprd02.prod.outlook.com>
X-TCD-Routed-via-EOP: Routed via EOP
X-TCD-ROUTED: Passed-Transport-Routing-Rules
X-MS-Oob-TLC-OOBClassifiers: OLM:229;
X-MS-Exchange-SenderADCheck: 1
X-MS-Exchange-AntiSpam-Relay: 0
X-Microsoft-Antispam: BCL:0;
X-Microsoft-Antispam-Message-Info: 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
X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:DB7PR02MB5113.eurprd02.prod.outlook.com; PTR:; CAT:NONE;  SFS:(4636009)(366004)(66476007)(66946007)(316002)(2616005)(786003)(66556008)(186003)(36756003)(83380400001)(31686004)(44832011)(235185007)(5660300002)(8676002)(45080400002)(508600001)(6512007)(2906002)(38100700002)(110136005)(86362001)(6506007)(53546011)(21480400003)(15650500001)(33964004)(6486002)(31696002)(8936002)(45980500001)(43740500002); DIR:OUT; SFP:1102; 
X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1
X-MS-Exchange-AntiSpam-MessageData-0: =?utf-8?B?c0tQajVqbzhFd3M1OUloWHNtaWowbStTMmNqYVQrclBmS3VSeU5MSFBFTVFQ?= =?utf-8?B?WWpIejU5TmNMNUFmaWNrY0k4Tnh4TXJ3b0dmNS9TYS9KZW1wZWJLcGF6Ylh3?= =?utf-8?B?aVI0b21sa01DTmpjd29ydVlISWhVRnkxNmo4QTBENGQralcra29pU3JXWWl3?= =?utf-8?B?eFgyK05nK1hIcm9EZUN1SWIxemdwdDlacjhqblVkcFNRZXoxdTYvOU9uQTlx?= =?utf-8?B?MmNrd2Vrdzh3U0pRQVhxRzRXQ2xVKzJaeFpoaHJHWnMwbkpzbWVNdDFvNXFE?= =?utf-8?B?YmdhOUpPcUhnMEV5emhXQll6REdpb2s1SDJjUWgzZXFLQ29IZzhxd0hvUGFG?= =?utf-8?B?dCtkUUV3eXZUak0zdkZUYnBXMldTMWdoZWorYm8xOTBiZDZVb1ExWTVwbURL?= =?utf-8?B?d1V6UGZWQWNqZkQyKzRFUjgwVUc1bkVoOE11WXVWYTBIL2dGRDN1WFhKaVNx?= =?utf-8?B?NFJrNTB1b1pOVEtyV21Ob0xsbTQxbDBmRDNjaGhaNUw2dXJHNFFKRlVTYmU3?= =?utf-8?B?RkF5NCsxbllubStQRSt6dnFNWnZWQzFBNEV5KzVIcU9Ta0JYRmFNcDh2ZGdi?= =?utf-8?B?SFRHcVY4NXJkYzNIQU91cHZrYktnN04xSVJ5Q1lJcmlBcTVWM1pKVkFLaUtI?= =?utf-8?B?S0JLWEJUMStXQ0FhRmU5bUpNelpxNDdDSDRzZGRlTm0xUEhyc2RaaFh0SVpO?= =?utf-8?B?ZXhKR1RtT3hJdkVXZm1ISjNubzB0T3REdjNBYnpCcVlmbHI3eFZmam1BQnFN?= =?utf-8?B?WExVSjVMSFB3elhaNEtBRVlLVnlzU3A0a1Z2R00rdE8vQlRDcEpaT0dJc2F2?= =?utf-8?B?cFNDeVRRRElqeFR1MDlkeDRHVEJ0QzlnSmp2aytyVzNTeVU5MTduTTkra2RD?= =?utf-8?B?V3ZRVXZiMlMwbEd2SGpoVWpKUllKUFpEeHh2eFlvZS96S0hkL0ZHRXg5YWk0?= =?utf-8?B?Wko5bFMxeW5ZdGN2R2hNSWRueURTWUpyNTBjVnJRQ05EN1dBdy9kaTNBREhR?= =?utf-8?B?QkN3RnZUVWpCbVVJMnZzSGgremJLZ2g1Umt0dXJhbTN2YmU3U1pENTMwRWlI?= =?utf-8?B?THBOcmZsTzRzcDc0QU0rVkpTbWpGelBoN3d6STVlT2lXSXNsZjFDMUZQYXVv?= =?utf-8?B?RU9RVk9xcVFtY0FhMjlYbWhyT3pYNWpYeDh6VVlpQjJKcEJXajgwMENoWjdj?= =?utf-8?B?VVJXdUFHSG5mR0hxeXNnb0YvU1hIWEhKbElmME1JYU84V0x0VkhCU0RDbmND?= =?utf-8?B?L3l5dzkzMitrRkdjV0tPaHAyMElQVEUvbGl3ZGY1dGdhOVVwTnB1KzdYMnNz?= =?utf-8?B?amcrcWFINmJkS3NYRG1sam5HQm40czhLMG5sdTc3Q1R5UHh2KzdTdVZ0ZndK?= =?utf-8?B?dXgrZUVRWWVURkdMbG1wSHNmYitFd2JIVVZ0eVJYV1RFcWRoRnd4aGV3dk1w?= =?utf-8?B?SzZWSTJuOXVhZ2VlV01BLzY1dHoyRHpnZ2h1TTRwNjlWUE1Pd3hWbnFWc04r?= =?utf-8?B?SVNhSi94MXFPcmkrOGVlS0N0VTkwdk5hTHYyTW1MOWJTT0NxdVhRckMxTWsx?= =?utf-8?B?NUljYm91dWc3RG5Ub2V2VGdTcEJkQTY2UTdEdWpWeTF6MWNqRSt6dytkUW9F?= =?utf-8?B?K21nOXpuamtNMEdJeWx1b3U1OVFLRDh3ZVhGYkpOMk5ENkNmNy9FZkJ0K1Vv?= =?utf-8?B?L0lTZUlFdVpIVXo4Ym9YTTJWVnBnRDhadWNVZ09seFlzeGl3ZTFhaVlJVFlm?= =?utf-8?B?ck9YRExaKzN6Nmd5T3JJZFZvRjZZWW1DYTlCZ0s1cmMzNUZjQmx3OFlnUU9i?= =?utf-8?B?Y21MZ2NKajJZSmNydWhIYVNsdmRweUE1R0hZcldJUm9WS2tLemVzRTYvSXlJ?= =?utf-8?B?MVBMM08xM3JnSmpmOXFLK284UGlrTkpDMGs4N3VXemtKU2RHMStHQkl1djNp?= =?utf-8?B?by9GM1RncVVlY2oxbU1ETWhxRVY0bjlNVERrRW5zOVgzYVcvV1pSM1l5alBq?= =?utf-8?B?eld1VlltdWhvK2hWeVhXc1dPRFdUbzQ1a2xrRnRKbFNSTGtrV3Z3SjE5WEhn?= =?utf-8?B?M3VjbHd0NWl4L1BQNHc0U3drbDVrUDFldXAxT0U4MERCSmZlNGx3UW5sdUw5?= =?utf-8?B?VThEdXdCT2pIZmQ5WDZJNDlpcTQ2R1Q1TXcvV1hMVDZzVUE2OTgrVDM5RnVI?= =?utf-8?Q?VAQfil29P+PB1uJJ2WIaRoE=3D?=
X-OriginatorOrg: cs.tcd.ie
X-MS-Exchange-CrossTenant-Network-Message-Id: 235e45e2-8283-48e8-070e-08d9d094b890
X-MS-Exchange-CrossTenant-AuthSource: DB7PR02MB5113.eurprd02.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 05 Jan 2022 21:45:42.2927 (UTC)
X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted
X-MS-Exchange-CrossTenant-Id: d595be8d-b306-45f4-8064-9e5b82fbe52b
X-MS-Exchange-CrossTenant-MailboxType: HOSTED
X-MS-Exchange-CrossTenant-UserPrincipalName: FvuyU1J74OcdJwLpw6GXsfPlKuSDbWYwAPlSV+6EIv/ncBvJvMNcvXahLSZTMSnZ
X-MS-Exchange-Transport-CrossTenantHeadersStamped: DB9PR02MB6652
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/PkaXk3ghiFt28C27l4stzBCkh9g>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 05 Jan 2022 21:45:52 -0000

--------------6iCS5SguaFTCuhCDKrELvYm0
Content-Type: multipart/mixed; boundary="------------fXnp8u4IrHTWC0kJEqK0eYfI";
 protected-headers="v1"
From: Stephen Farrell <stephen.farrell@cs.tcd.ie>
To: Roman Danyliw <rdd@cert.org>, "ppm@ietf.org" <ppm@ietf.org>
Message-ID: <9d699c4f-3b00-ab9a-f7bd-bfac507edc75@cs.tcd.ie>
Subject: Re: [Ppm] Call for consensus on updated charter
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>

--------------fXnp8u4IrHTWC0kJEqK0eYfI
Content-Type: multipart/mixed; boundary="------------z1sqVouJjfn0SRtc3g6q4k4d"

--------------z1sqVouJjfn0SRtc3g6q4k4d
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: base64
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--------------z1sqVouJjfn0SRtc3g6q4k4d
Content-Type: application/pgp-keys; name="OpenPGP_0x5AB2FAF17B172BEA.asc"
Content-Disposition: attachment; filename="OpenPGP_0x5AB2FAF17B172BEA.asc"
Content-Description: OpenPGP public key
Content-Transfer-Encoding: quoted-printable

-----BEGIN PGP PUBLIC KEY BLOCK-----
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=3D40Nd
-----END PGP PUBLIC KEY BLOCK-----
--------------z1sqVouJjfn0SRtc3g6q4k4d--


--------------fXnp8u4IrHTWC0kJEqK0eYfI--

--------------6iCS5SguaFTCuhCDKrELvYm0
Content-Type: application/pgp-signature; name="OpenPGP_signature.asc"
Content-Description: OpenPGP digital signature
Content-Disposition: attachment; filename="OpenPGP_signature"

-----BEGIN PGP SIGNATURE-----
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=LK1o
-----END PGP SIGNATURE-----

--------------6iCS5SguaFTCuhCDKrELvYm0--


From nobody Wed Jan  5 14:37:58 2022
Return-Path: <timg@letsencrypt.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 4FC903A0964 for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 14:37:49 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.674
X-Spam-Level: 
X-Spam-Status: No, score=-2.674 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.576, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=letsencrypt.org
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id bdsAV_0uKF4E for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 14:37:44 -0800 (PST)
Received: from mail-qt1-x832.google.com (mail-qt1-x832.google.com [IPv6:2607:f8b0:4864:20::832]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 1E92A3A07F0 for <ppm@ietf.org>; Wed,  5 Jan 2022 14:37:27 -0800 (PST)
Received: by mail-qt1-x832.google.com with SMTP id p19so481983qtw.12 for <ppm@ietf.org>; Wed, 05 Jan 2022 14:37:27 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=letsencrypt.org; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=M9q/fRjA8nxP3rKGuijUQkpo15cXLJQorDLWHslUtJY=; b=G8p/b+NvPMGd2K/PG2bBtlEQ3FXYaQam+PALm3M1hXctCaV+aG37uGaEMkZJM/XEUP ZI+QrePycMGVut3lX/+WIwrudPd1OQGCCl2qPJ0PEuAY/2mBdsbUrl8t/4vWyPckddn7 Fk1AR16oAFyd5jnOOttUWZom8/LnLp6KiSRKQ=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=M9q/fRjA8nxP3rKGuijUQkpo15cXLJQorDLWHslUtJY=; b=UyzX6/aP9QzgUPaiXQwII5x7vm8zZdspAppBNRVWO6sM6x6pQ1HWG6sRzbIH2XEFBL jvzseoNLsFkUCiMwVzayjoxu+nPpEWMGoxKCBr1s58Tq73mI+JGsijrjATAjdI0/vMVn UcmgrFzKU+rJApjDYujikb4mfU/FN/rZnHs6kDy1BUdYruNBWt4ZXRdvepEt5BjHyTin LmBcmHqIJ+uidyH8iLfkD+/W0eEYL1LcRtb7rzwlP5sgCehmasC1I1NzTYY9gXLLsrVY CloE5s/Oj3Z5GMKkT60KHGwNRvbiw5ly+pMJaEPqpTw8/m45xGwDEtuiQJmiIp6oli8F 3TYA==
X-Gm-Message-State: AOAM5328OxtbrD4tLratu/11lY/Q3wv8uENvKY9JkOlxQLy5Zs1QAmQE IcAotzxAJPO2ZvtglITKrpJJtqXgKcaSlywR1S9MfFwK17g=
X-Google-Smtp-Source: ABdhPJy6+6I29OIJ2/f8hhcx7s94hD2Vtd8VwqOQE1OMTmJ94Pu9cbYkSB2eoGD52v3X3OoaU6ZfMGBeOEpcg2V/rcQ=
X-Received: by 2002:a05:622a:15cc:: with SMTP id d12mr495895qty.190.1641422244941;  Wed, 05 Jan 2022 14:37:24 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Tim Geoghegan <timg@letsencrypt.org>
Date: Wed, 5 Jan 2022 14:37:14 -0800
Message-ID: <CABN231rHSq0S_R79PkznpRfnYK4saecVBEHR2r47tUh10+ipSA@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="000000000000d6268105d4dd6599"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/hyVWRA5Yu3hNNlRAIyyCzLoVKCY>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 05 Jan 2022 22:37:57 -0000

--000000000000d6268105d4dd6599
Content-Type: text/plain; charset="UTF-8"

On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw <rdd@cert.org> wrote:

> (1) Do you support the charter text [3]? Or do you have objections or
> blocking concerns (please describe what they might be)?
> Yes
> If you do support the charter text:
> (2) Are you willing to author or participate in the developed of the WG
> drafts?
>
Yes

> (3) Are you willing to review the WG drafts?
>
Yes

> (4) Are you interested in implementing the WG drafts?
>
Yes

Tim Geoghegan
ISRG


> This call for consensus will end on Thursday, January 20.
>
> Regards,
> Roman
>
> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> [2] charter-ietf-priv-00-00 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-00/
> [3] charter-ietf-priv-00-01 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-01/
> [4] diff of 00-00 vs. 00-01 --
> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt
> [5] https://datatracker.ietf.org/wg/priv/about/
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000d6268105d4dd6599
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail=
_attr">On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw &lt;<a href=3D"mailto:r=
dd@cert.org">rdd@cert.org</a>&gt; wrote:<br></div><blockquote class=3D"gmai=
l_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,20=
4,204);padding-left:1ex">(1) Do you support the charter text [3]? Or do you=
 have objections or blocking concerns (please describe what they might be)?=
<br>
Yes<br>
If you do support the charter text:<br>
(2) Are you willing to author or participate in the developed of the WG dra=
fts?<br></blockquote><div>Yes <br></div><blockquote class=3D"gmail_quote" s=
tyle=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);pad=
ding-left:1ex">
(3) Are you willing to review the WG drafts?<br></blockquote><div>Yes <br><=
/div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bo=
rder-left:1px solid rgb(204,204,204);padding-left:1ex">
(4) Are you interested in implementing the WG drafts?<br></blockquote><div>=
Yes</div><div><br></div><div>Tim Geoghegan</div><div>ISRG</div><div><br></d=
iv><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bord=
er-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
This call for consensus will end on Thursday, January 20.<br>
<br>
Regards,<br>
Roman <br>
<br>
[1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=3D"n=
oreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-112-p=
riv/</a><br>
[2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-00/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-00/</a><br>
[3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-01/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-01/</a><br>
[4] diff of 00-00 vs. 00-01 -- <a href=3D"https://www.ietf.org/rfcdiff?url1=
=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmiles=
tones-00-00.txt&amp;url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fchart=
er-ietf-priv%2Fwithmilestones-00-01.txt" rel=3D"noreferrer" target=3D"_blan=
k">https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2=
Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&amp;url2=3Dhttps%3A%2F%=
2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt=
</a><br>
[5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"noreferr=
er" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div>

--000000000000d6268105d4dd6599--


From nobody Wed Jan  5 16:26:27 2022
Return-Path: <caw@heapingbits.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 1DFBD3A0E08 for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 16:26:26 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.098
X-Spam-Level: 
X-Spam-Status: No, score=-2.098 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=heapingbits.net header.b=M/7RTMei; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=eoO8XJbI
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 15dvyEC0ZFDg for <ppm@ietfa.amsl.com>; Wed,  5 Jan 2022 16:26:20 -0800 (PST)
Received: from wout2-smtp.messagingengine.com (wout2-smtp.messagingengine.com [64.147.123.25]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id E116A3A0E07 for <ppm@ietf.org>; Wed,  5 Jan 2022 16:26:20 -0800 (PST)
Received: from compute6.internal (compute6.nyi.internal [10.202.2.46]) by mailout.west.internal (Postfix) with ESMTP id 3F0673200657; Wed,  5 Jan 2022 19:26:20 -0500 (EST)
Received: from mailfrontend2 ([10.202.2.163]) by compute6.internal (MEProxy); Wed, 05 Jan 2022 19:26:20 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=heapingbits.net; h=content-type:mime-version:subject:from:in-reply-to:date:cc :content-transfer-encoding:message-id:references:to; s=fm3; bh=x VhOidiXJb4kGA+3Wko924HbGHZbaEcR8Lv0z2kxbjg=; b=M/7RTMeiQjoym1gPi z+GVUjRIsTjlthyB+bB5Bjpxz9LSOyrIyWmrBsPKYbI4t966oUNx+AWu1Jx7w4++ zxSMe7OEtkB5KMasRRs//exLcha05SYQiQYeMYh6LfpjDpkfFbXaXg3aalySlHdk tOoYn3BT/hOL94vG3lLjjsWqfPBgNItumWxw0A4t5s4pnspXlRSOF+cmlguoy6ve XyXzoL56HNNDNkzhhbNXBXcojp65zgCmdvn06OoHZ+AG2cR32wt7Dz2cnrvhg+Fd DGCq0wAgQgtJpRN1e8bBHNtNoFaad8GrT2hBix4wI23iNBIoQztZSJ5rQZHx2mQO yv9fA==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-transfer-encoding:content-type :date:from:in-reply-to:message-id:mime-version:references :subject:to:x-me-proxy:x-me-proxy:x-me-sender:x-me-sender :x-sasl-enc; s=fm1; bh=xVhOidiXJb4kGA+3Wko924HbGHZbaEcR8Lv0z2kxb jg=; b=eoO8XJbIuA0ERunIZP1C0Pi9ShlZvMbCsZ6Bd6ZghJ0/i2z7lneqm0bVq YIP6pM9Z8WsT8XSktw2IUCUswfHYlylztegaS6NS1aI0ZFsP30qVCLz+XdERArCr a1sACQr4w0C7UGpJq1sz6Qi8izsACejVwGaowbpN3/PTGZz72gIBW9i8Mvu7XILJ Fex9Cbzst5rqQRokMquG7+8wYXTqgbFCuTu2+JONYOVCxgnytlSDitD28hPODPYW xEneKcx6jo76bPr/e8HuYUv8Ews4VMzCeyVIlCm3yMEpvi5NaLfucQcqrFDk3iXO CbEQfhaTVWXgVUtXZ1Qwhge/Hpv+A==
X-ME-Sender: <xms:KzfWYUM0UwJzJqAeJEeZ18bVg-Nb2pQ_gmF2XSs-OVCjFFcszMcVFQ> <xme:KzfWYa-OG05tybdFGS6VmkBix-Yzh0bwuXgiKKkxQz8o1icbN-Xhhg95HljYwXUxp oWwc4iS-mLHot8Y0Kw>
X-ME-Received: <xmr:KzfWYbRSoUbmihu0b8abpqUGXqJv8g5J1NgnwjNtxUqWGlK4zO6ycFsZP7QAo25blQbbS67-kwQJMh54xDDcv4MwFx4XoV88qg>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvuddrudefjedguddttdcutefuodetggdotefrod ftvfcurfhrohhfihhlvgemucfhrghsthforghilhdpqfgfvfdpuffrtefokffrpgfnqfgh necuuegrihhlohhuthemuceftddtnecunecujfgurheptggguffhjgffgffkfhfvofesth hqmhdthhdtvdenucfhrhhomhepvehhrhhishhtohhphhgvrhcuhghoohguuceotggrfies hhgvrghpihhnghgsihhtshdrnhgvtheqnecuggftrfgrthhtvghrnhepheeihfejveekie dvvdehffevtdetfedtvefgjeeujefgieeijedukedutdelgeeknecuvehluhhsthgvrhfu ihiivgeptdenucfrrghrrghmpehmrghilhhfrhhomheptggrfieshhgvrghpihhnghgsih htshdrnhgvth
X-ME-Proxy: <xmx:KzfWYcta0Z3CO0V0aU9tydcD0ML0cH6acZkkXPb-fUmadDvWMpyu2A> <xmx:KzfWYcfOHUPvk6jqfudHDZrr7-Bx5Hm5I5KwRKbND86cK7RVSQMi8A> <xmx:KzfWYQ1x0g2ioZ2ZJz63jqazvjh3ze6BDTQJv27sw7zHPizl7R3u_A> <xmx:KzfWYUmSejVLppc_28fNwQY4A2JA250b96rMx9j3SaxbsaARmX2sLA>
Received: by mail.messagingengine.com (Postfix) with ESMTPA; Wed, 5 Jan 2022 19:26:19 -0500 (EST)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 14.0 \(3654.120.0.1.13\))
From: Christopher Wood <caw@heapingbits.net>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Date: Wed, 5 Jan 2022 16:26:14 -0800
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <2D446D57-8CA2-4BF7-95DB-303AE52A8D9C@heapingbits.net>
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
To: Roman Danyliw <rdd@cert.org>
X-Mailer: Apple Mail (2.3654.120.0.1.13)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/BrHun4y0ShPt6ngY9b52dCHWV78>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 06 Jan 2022 00:26:26 -0000

> On Jan 5, 2022, at 1:19 PM, Roman Danyliw <rdd@cert.org> wrote:
>=20
> Hi!
>=20
> Happy New Year! Thanks for the robust discussion at the PRIV BOF at =
IETF 112.  The meeting notes [1] captured four major themes of feedback =
based on the presented charter [2] (version 00-00).  Specifically:
>=20
> =3D=3D[ snip from meeting notes ]=3D=3D
> (a) Generalize the text to make it clear that flexibility for =
alternatives and additions to approach is possible (i.e., more than =
Prio, heavy hitters)
> (b) Refine the definition of aggregations so as to not restrict =
alternatives
> (c) Add a work item(s) to document the abuse cases and associated =
mitigations from the perspective of all participants in the architecture
> (d) Revision of the WG name to convey a more narrow scope
> =3D=3D[ snip ]=3D=3D =20
>=20
> In response to this feedback, the proposed charter text has been =
revised at [3] (version 00-01).  Also, see [4] for a diff of this =
revision against the original text presented at IETF 112.  For =
convenience, this charter text is loaded into the datatracker using the =
BOF name (PRIV). As there wasn't consensus on "PRIV" as the WG name per =
feedback (d) or the now proposed alternative of "PPM", it didn't make =
sense to ask the Secretariat to create a new entry in the datatracker =
yet.  If consensus is found, a new datatracker entry will be made prior =
to sending the charter to the IESG.
>=20
> This email is intended to start a thread to gauge interest in forming =
a WG around this new, refined charter language [3].  Please respond to =
the list:
>=20
> (1) Do you support the charter text [3]? Or do you have objections or =
blocking concerns (please describe what they might be)?

Yes.
> If you do support the charter text:
> (2) Are you willing to author or participate in the developed of the =
WG drafts?
Yes.
> (3) Are you willing to review the WG drafts?
Yes.
> (4) Are you interested in implementing the WG drafts?
Yes.

Best,
Chris=


From nobody Fri Jan  7 08:01:58 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id CB7103A0A51 for <ppm@ietfa.amsl.com>; Fri,  7 Jan 2022 08:01:56 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 9S1U3kYOyzAN for <ppm@ietfa.amsl.com>; Fri,  7 Jan 2022 08:01:54 -0800 (PST)
Received: from mail-pf1-x429.google.com (mail-pf1-x429.google.com [IPv6:2607:f8b0:4864:20::429]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 57B173A0A42 for <ppm@ietf.org>; Fri,  7 Jan 2022 08:01:54 -0800 (PST)
Received: by mail-pf1-x429.google.com with SMTP id m1so5494076pfk.8 for <ppm@ietf.org>; Fri, 07 Jan 2022 08:01:54 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=pineL6fKKQV+bMZWFje5vo+2us+dWGpaKpLgT/va8T0=; b=x/TGaCe5kL/sXpdwYnVgkRCzukFryr/pRRVU5gRDMEBmtZRFg14PVYAl8A/p3bRHPE 6J3yBJINF8P7OuOxXo0IT20TVslMG73Z/6yQ/EsE8/MsfD+7rc+gErGseKHgcgKoHjF1 WuPk0RFAjYAXGiCW/5a9l6vPYsQDX6BfXWxPY=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=pineL6fKKQV+bMZWFje5vo+2us+dWGpaKpLgT/va8T0=; b=w7Pa/2ObaXZ6h+n+w+FezT/ktpYRY/QefeAiELHB54w8f8KFtPpCMx3adexUx9A0M2 T+3x7Xpbh+RoeB2doeQV+gEuYeUHqcGolOTnvMc4SNw2MFOR2Xh+AKIYAL+SnOvjSa5K hLJUCDTdeB1cQR/DJuOV0n4My1UOo53TrDPsgC3GPYGcUicuwIKxm2gkwxXmuUkq+yHA wIeGgu9fFZ53bEG0KDTHm8mUIE0BRjTMVxCpkdLiMo4qCepjxJDkzyPfkCEHNpRTeZ/7 Esx0k4P5pkaYuXoQbdduOpPxeEvI9VBX0Rl8zZyXVcSMljXH4Rtdz0PrZsgpziCsSqsA +epA==
X-Gm-Message-State: AOAM5339lMscxezyW2qm4OX2HkNp8w9YfbOvBUGIjZt1j1ID2pjRhoHJ gtRbuWl+9u+zy8pq9CO6YTUY9R83SzPdBhjCmaUuTwblkNrmyA==
X-Google-Smtp-Source: ABdhPJw/sIsF+PKlkUUtuVk9EfUVyzVm7PDxtWz1313n5pSCDzWl4N+61NmvelFrR8rxT81F7MNmJTnz3rh9pr0ykSo=
X-Received: by 2002:a05:6a00:174b:b0:4ba:c9f4:d4ee with SMTP id j11-20020a056a00174b00b004bac9f4d4eemr65264828pfc.3.1641571312671; Fri, 07 Jan 2022 08:01:52 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Fri, 7 Jan 2022 08:01:41 -0800
Message-ID: <CAG2Zi22sJ1n_nKgP0e4R+jxr4ObvL7jT7zLqhKFpZWoOkapfZA@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="000000000000f7477505d5001a13"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/l54YMPvG6LK4_yLUu3DYcHDVXss>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 07 Jan 2022 16:01:57 -0000

--000000000000f7477505d5001a13
Content-Type: text/plain; charset="UTF-8"

(1) Yes
(2) Yes
(3) Yes
(4) Yes

Chris P.

On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw <rdd@cert.org> wrote:

> Hi!
>
> Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF
> 112.  The meeting notes [1] captured four major themes of feedback based on
> the presented charter [2] (version 00-00).  Specifically:
>
> ==[ snip from meeting notes ]==
> (a) Generalize the text to make it clear that flexibility for alternatives
> and additions to approach is possible (i.e., more than Prio, heavy hitters)
> (b) Refine the definition of aggregations so as to not restrict
> alternatives
> (c) Add a work item(s) to document the abuse cases and associated
> mitigations from the perspective of all participants in the architecture
> (d) Revision of the WG name to convey a more narrow scope
> ==[ snip ]==
>
> In response to this feedback, the proposed charter text has been revised
> at [3] (version 00-01).  Also, see [4] for a diff of this revision against
> the original text presented at IETF 112.  For convenience, this charter
> text is loaded into the datatracker using the BOF name (PRIV). As there
> wasn't consensus on "PRIV" as the WG name per feedback (d) or the now
> proposed alternative of "PPM", it didn't make sense to ask the Secretariat
> to create a new entry in the datatracker yet.  If consensus is found, a new
> datatracker entry will be made prior to sending the charter to the IESG.
>
> This email is intended to start a thread to gauge interest in forming a WG
> around this new, refined charter language [3].  Please respond to the list:
>
> (1) Do you support the charter text [3]? Or do you have objections or
> blocking concerns (please describe what they might be)?
>
> If you do support the charter text:
> (2) Are you willing to author or participate in the developed of the WG
> drafts?
> (3) Are you willing to review the WG drafts?
> (4) Are you interested in implementing the WG drafts?
>
> This call for consensus will end on Thursday, January 20.
>
> Regards,
> Roman
>
> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> [2] charter-ietf-priv-00-00 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-00/
> [3] charter-ietf-priv-00-01 -
> https://datatracker.ietf.org/doc/charter-ietf-priv/00-01/
> [4] diff of 00-00 vs. 00-01 --
> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt
> [5] https://datatracker.ietf.org/wg/priv/about/
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000f7477505d5001a13
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>(1) Yes<br></div><div>(2) Yes</div><div>(3) Yes</div>=
<div>(4) Yes</div><div><br></div><div>Chris P.</div></div><br><div class=3D=
"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 5, 2022 at =
1:19 PM Roman Danyliw &lt;<a href=3D"mailto:rdd@cert.org">rdd@cert.org</a>&=
gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0=
px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi!<b=
r>
<br>
Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF 11=
2.=C2=A0 The meeting notes [1] captured four major themes of feedback based=
 on the presented charter [2] (version 00-00).=C2=A0 Specifically:<br>
<br>
=3D=3D[ snip from meeting notes ]=3D=3D<br>
(a) Generalize the text to make it clear that flexibility for alternatives =
and additions to approach is possible (i.e., more than Prio, heavy hitters)=
<br>
(b) Refine the definition of aggregations so as to not restrict alternative=
s<br>
(c) Add a work item(s) to document the abuse cases and associated mitigatio=
ns from the perspective of all participants in the architecture<br>
(d) Revision of the WG name to convey a more narrow scope<br>
=3D=3D[ snip ]=3D=3D=C2=A0 <br>
<br>
In response to this feedback, the proposed charter text has been revised at=
 [3] (version 00-01).=C2=A0 Also, see [4] for a diff of this revision again=
st the original text presented at IETF 112.=C2=A0 For convenience, this cha=
rter text is loaded into the datatracker using the BOF name (PRIV). As ther=
e wasn&#39;t consensus on &quot;PRIV&quot; as the WG name per feedback (d) =
or the now proposed alternative of &quot;PPM&quot;, it didn&#39;t make sens=
e to ask the Secretariat to create a new entry in the datatracker yet.=C2=
=A0 If consensus is found, a new datatracker entry will be made prior to se=
nding the charter to the IESG.<br>
<br>
This email is intended to start a thread to gauge interest in forming a WG =
around this new, refined charter language [3].=C2=A0 Please respond to the =
list:<br>
<br>
(1) Do you support the charter text [3]? Or do you have objections or block=
ing concerns (please describe what they might be)?<br>
<br>
If you do support the charter text:<br>
(2) Are you willing to author or participate in the developed of the WG dra=
fts?<br>
(3) Are you willing to review the WG drafts?<br>
(4) Are you interested in implementing the WG drafts?<br>
<br>
This call for consensus will end on Thursday, January 20.<br>
<br>
Regards,<br>
Roman <br>
<br>
[1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=3D"n=
oreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-112-p=
riv/</a><br>
[2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-00/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-00/</a><br>
[3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-01/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-01/</a><br>
[4] diff of 00-00 vs. 00-01 -- <a href=3D"https://www.ietf.org/rfcdiff?url1=
=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmiles=
tones-00-00.txt&amp;url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fchart=
er-ietf-priv%2Fwithmilestones-00-01.txt" rel=3D"noreferrer" target=3D"_blan=
k">https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2=
Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&amp;url2=3Dhttps%3A%2F%=
2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt=
</a><br>
[5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"noreferr=
er" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--000000000000f7477505d5001a13--


From nobody Fri Jan  7 08:31:11 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 90BCA3A0B6C for <ppm@ietfa.amsl.com>; Fri,  7 Jan 2022 08:31:09 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 1pIXU6-7LaW7 for <ppm@ietfa.amsl.com>; Fri,  7 Jan 2022 08:31:05 -0800 (PST)
Received: from mail-lf1-x12b.google.com (mail-lf1-x12b.google.com [IPv6:2a00:1450:4864:20::12b]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 18CB03A0B6B for <ppm@ietf.org>; Fri,  7 Jan 2022 08:31:05 -0800 (PST)
Received: by mail-lf1-x12b.google.com with SMTP id x7so17409061lfu.8 for <ppm@ietf.org>; Fri, 07 Jan 2022 08:31:04 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to;  bh=t+x1QVE3Wdq4riWY0wQjPsLklspkk7MOD6/f4KmjAzc=; b=Xtv6sfeV3WfoUaAe83bGoQo7CGhSlW0IUUtQvvWLn+ojwzBKsU3QjLD1rDqf5TgfNO li1ygOy5aivpimxtRKSTlc+4wQBOiiowWC3CGoXZZLH0PgsDCyKZO5eceK5eKCYEhAAO xF8nNb4yB23A+NIXssulFZ5HHzDqesHicPKnNcvyfd6PliXqZR0GDzxheYNW1WCNSPkm +MmzAEIk+t66LaTzS7FJBH08eiR5+1p5JxTWbFQ63KItJCUIyf+DwgYFO4yzcMFmbJ72 SY66GKh/v0N8B3xYRVomoXhirOXoxwBqiMPXhO8LUUKWTPuLlIPyuA5RxYeYPW5B2yen Kqiw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to; bh=t+x1QVE3Wdq4riWY0wQjPsLklspkk7MOD6/f4KmjAzc=; b=OrnYIpGnw0snHhYHAzTLNDkYoCcmbxDr2t2NV8880GfSlZ/TF5RKnZOwjwkzFj3/ho Pk9SUMhtD/QVBXKmIAtD/rlHLeol7OY8XYS1z6HhzW/4JlXVsthbaxCrGAi5pEjW4GOg S0GirHk/UM/nJ+YI/CVTIIuw8xJVmgZOYB8S7vzASYLOGYVSvTX0Te87f/kjXd0iodLK 5ogMYiO83Pnna++U1TRWKITqbAyxgzW+u64SiUEVfcaplIowiwGoG6D8oJLKygBpihaQ J9JbMy2AEUP1X5P/p47asbbTnENbkpffyT+oHmdZVY48Tx+yYy3Er6LYSVEe0vtCxxHa ulog==
X-Gm-Message-State: AOAM530ync3563CtG5UjwsEGydrbw27KY+tsDWkZZFwEEOMrq3ZfDBjN dnh0nyCTrV97LKHA18Lzfq5InTNGb1aX3NUtB+td54dw
X-Google-Smtp-Source: ABdhPJxLwy5eUrXyLH+fasqqssnpO4tB6Jc9JpzpfiCztcXtw39NkddPdUxTenUgCPoSQpO9+INWtr7pgkqrbBz5o0E=
X-Received: by 2002:a05:6512:3082:: with SMTP id z2mr52942141lfd.261.1641573061908;  Fri, 07 Jan 2022 08:31:01 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <CAG2Zi22sJ1n_nKgP0e4R+jxr4ObvL7jT7zLqhKFpZWoOkapfZA@mail.gmail.com>
In-Reply-To: <CAG2Zi22sJ1n_nKgP0e4R+jxr4ObvL7jT7zLqhKFpZWoOkapfZA@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Fri, 7 Jan 2022 08:30:50 -0800
Message-ID: <CANzuiCj7b8iX55xyPtjJAdi6bKKCiHM4g5LL3-3jorm7m-b8jg@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>, "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="0000000000003a5cbf05d50083d5"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/0nhfi1INTv-CaNR-PwlINRzHni4>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 07 Jan 2022 16:31:10 -0000

--0000000000003a5cbf05d50083d5
Content-Type: text/plain; charset="UTF-8"

(1) Yes
(2) Yes
(3) Yes
(4) Possibly, pending commitment and resourcing.

Erik Taubeneck

On Fri, Jan 7, 2022 at 8:02 AM Christopher Patton <cpatton=
40cloudflare.com@dmarc.ietf.org> wrote:

> (1) Yes
> (2) Yes
> (3) Yes
> (4) Yes
>
> Chris P.
>
> On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw <rdd@cert.org> wrote:
>
>> Hi!
>>
>> Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF
>> 112.  The meeting notes [1] captured four major themes of feedback based on
>> the presented charter [2] (version 00-00).  Specifically:
>>
>> ==[ snip from meeting notes ]==
>> (a) Generalize the text to make it clear that flexibility for
>> alternatives and additions to approach is possible (i.e., more than Prio,
>> heavy hitters)
>> (b) Refine the definition of aggregations so as to not restrict
>> alternatives
>> (c) Add a work item(s) to document the abuse cases and associated
>> mitigations from the perspective of all participants in the architecture
>> (d) Revision of the WG name to convey a more narrow scope
>> ==[ snip ]==
>>
>> In response to this feedback, the proposed charter text has been revised
>> at [3] (version 00-01).  Also, see [4] for a diff of this revision against
>> the original text presented at IETF 112.  For convenience, this charter
>> text is loaded into the datatracker using the BOF name (PRIV). As there
>> wasn't consensus on "PRIV" as the WG name per feedback (d) or the now
>> proposed alternative of "PPM", it didn't make sense to ask the Secretariat
>> to create a new entry in the datatracker yet.  If consensus is found, a new
>> datatracker entry will be made prior to sending the charter to the IESG.
>>
>> This email is intended to start a thread to gauge interest in forming a
>> WG around this new, refined charter language [3].  Please respond to the
>> list:
>>
>> (1) Do you support the charter text [3]? Or do you have objections or
>> blocking concerns (please describe what they might be)?
>>
>> If you do support the charter text:
>> (2) Are you willing to author or participate in the developed of the WG
>> drafts?
>> (3) Are you willing to review the WG drafts?
>> (4) Are you interested in implementing the WG drafts?
>>
>> This call for consensus will end on Thursday, January 20.
>>
>> Regards,
>> Roman
>>
>> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
>> [2] charter-ietf-priv-00-00 -
>> https://datatracker.ietf.org/doc/charter-ietf-priv/00-00/
>> [3] charter-ietf-priv-00-01 -
>> https://datatracker.ietf.org/doc/charter-ietf-priv/00-01/
>> [4] diff of 00-00 vs. 00-01 --
>> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt
>> [5] https://datatracker.ietf.org/wg/priv/about/
>>
>> --
>> Ppm mailing list
>> Ppm@ietf.org
>> https://www.ietf.org/mailman/listinfo/ppm
>>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--0000000000003a5cbf05d50083d5
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">(1) Yes<div>(2) Yes</div><div>(3) Yes</div><div>(4) Possib=
ly, pending commitment and resourcing.</div><div><br></div><div>Erik Tauben=
eck</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Fri, Jan 7, 2022 at 8:02 AM Christopher Patton &lt;cpatton=3D<a=
 href=3D"mailto:40cloudflare.com@dmarc.ietf.org">40cloudflare.com@dmarc.iet=
f.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:=
1ex"><div dir=3D"ltr"><div>(1) Yes<br></div><div>(2) Yes</div><div>(3) Yes<=
/div><div>(4) Yes</div><div><br></div><div>Chris P.</div></div><br><div cla=
ss=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 5, 202=
2 at 1:19 PM Roman Danyliw &lt;<a href=3D"mailto:rdd@cert.org" target=3D"_b=
lank">rdd@cert.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote=
" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);=
padding-left:1ex">Hi!<br>
<br>
Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF 11=
2.=C2=A0 The meeting notes [1] captured four major themes of feedback based=
 on the presented charter [2] (version 00-00).=C2=A0 Specifically:<br>
<br>
=3D=3D[ snip from meeting notes ]=3D=3D<br>
(a) Generalize the text to make it clear that flexibility for alternatives =
and additions to approach is possible (i.e., more than Prio, heavy hitters)=
<br>
(b) Refine the definition of aggregations so as to not restrict alternative=
s<br>
(c) Add a work item(s) to document the abuse cases and associated mitigatio=
ns from the perspective of all participants in the architecture<br>
(d) Revision of the WG name to convey a more narrow scope<br>
=3D=3D[ snip ]=3D=3D=C2=A0 <br>
<br>
In response to this feedback, the proposed charter text has been revised at=
 [3] (version 00-01).=C2=A0 Also, see [4] for a diff of this revision again=
st the original text presented at IETF 112.=C2=A0 For convenience, this cha=
rter text is loaded into the datatracker using the BOF name (PRIV). As ther=
e wasn&#39;t consensus on &quot;PRIV&quot; as the WG name per feedback (d) =
or the now proposed alternative of &quot;PPM&quot;, it didn&#39;t make sens=
e to ask the Secretariat to create a new entry in the datatracker yet.=C2=
=A0 If consensus is found, a new datatracker entry will be made prior to se=
nding the charter to the IESG.<br>
<br>
This email is intended to start a thread to gauge interest in forming a WG =
around this new, refined charter language [3].=C2=A0 Please respond to the =
list:<br>
<br>
(1) Do you support the charter text [3]? Or do you have objections or block=
ing concerns (please describe what they might be)?<br>
<br>
If you do support the charter text:<br>
(2) Are you willing to author or participate in the developed of the WG dra=
fts?<br>
(3) Are you willing to review the WG drafts?<br>
(4) Are you interested in implementing the WG drafts?<br>
<br>
This call for consensus will end on Thursday, January 20.<br>
<br>
Regards,<br>
Roman <br>
<br>
[1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=3D"n=
oreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-112-p=
riv/</a><br>
[2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-00/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-00/</a><br>
[3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-01/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-01/</a><br>
[4] diff of 00-00 vs. 00-01 -- <a href=3D"https://www.ietf.org/rfcdiff?url1=
=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmiles=
tones-00-00.txt&amp;url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fchart=
er-ietf-priv%2Fwithmilestones-00-01.txt" rel=3D"noreferrer" target=3D"_blan=
k">https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2=
Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&amp;url2=3Dhttps%3A%2F%=
2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt=
</a><br>
[5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"noreferr=
er" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--0000000000003a5cbf05d50083d5--


From nobody Sat Jan  8 04:13:29 2022
Return-Path: <schoppmann@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 1D1F93A143C for <ppm@ietfa.amsl.com>; Sat,  8 Jan 2022 04:13:28 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 5IvBMWaXnxII for <ppm@ietfa.amsl.com>; Sat,  8 Jan 2022 04:13:23 -0800 (PST)
Received: from mail-wm1-x32b.google.com (mail-wm1-x32b.google.com [IPv6:2a00:1450:4864:20::32b]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 4ED713A143A for <ppm@ietf.org>; Sat,  8 Jan 2022 04:13:23 -0800 (PST)
Received: by mail-wm1-x32b.google.com with SMTP id l12-20020a7bc34c000000b003467c58cbdfso6424807wmj.2 for <ppm@ietf.org>; Sat, 08 Jan 2022 04:13:23 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=zeGn1ipkxuXvc+YRhnAOGMxka/ppwG/1idzMBHyiySU=; b=ALaHTQS14yc497+2fRvmDjQA7pzAAHiJ9Oz/AzgZ6el/Rs49iWgUN5DwFAT5Q2j4Mo jlz1oJcKS+HLggOgOTIndVmT1F/5nai5nilLTIbXHIhj15vguf4uYb+2dAMnR6KZTc6d MbHQ2lpiI5/NyibFtImLXBr91cK/aop9SL2TGsa2QUg2RVIDRPKGgHwl1Dkh2BKBalbz I7C9Wqjf0dgX218XBYKixqe3YV2UE1NbiyMEyHo/YaD8wqeUBquiOYL83oUz8nb0k2vQ KipGGe4dnepy2txhA34MKGKaHRB2/fHm6XEva1Oo0/SKR1yfCj/560jII2IDL7eAkQiE 7W/A==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=zeGn1ipkxuXvc+YRhnAOGMxka/ppwG/1idzMBHyiySU=; b=UmCzUTnsI2ZAlP4Itp+zS9F9gdnhoD5X7G/8fzb0QKJ63AB2kY2tOOsM5m0RfswAN2 +xCF7cTGvWz0zS0wWLWgiM5n6sHKOQu/Zx1dS93Ha0i1T6AiDufq3MK+WscIuI63friz /8RiDC2u2OE7w+f9lECCETTBJMcPftKzYLHjWwr8j21XYodDeuic7xfIoY2jl5LJZKqP bmBg8L+CdAwQzOAjBlsibpGj2DFL0mSCoSypqJXjOhZi0/HOAI8BJOrfvl/9Z+jFGd5S yKcq0WhzDArr/YdrmEHyQJQU5u29AL0p8jRQrTTDo/i/J9aIis7r5Iy+ep/+ZLx0lFe7 wGlA==
X-Gm-Message-State: AOAM531lsHR3KgEGyOPn1qLYdVqdfZ8uKrM6gKB5SCBywY/Cnl0a0mNO g7SN+IjnsRv9Yx6Zo4qDOj6m0lVEjio3EvtPd3VBZQ==
X-Google-Smtp-Source: ABdhPJwVwSIZdtuIvNLdADo8DhrKjrgDweNLf1JR8R5JgzteeEmPzkhu9v7tpnwZpw/s9hlXTA0+UqYVOFfvGCAi68A=
X-Received: by 2002:a1c:a747:: with SMTP id q68mr13135121wme.98.1641643996578;  Sat, 08 Jan 2022 04:13:16 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <CAG2Zi22sJ1n_nKgP0e4R+jxr4ObvL7jT7zLqhKFpZWoOkapfZA@mail.gmail.com> <CANzuiCj7b8iX55xyPtjJAdi6bKKCiHM4g5LL3-3jorm7m-b8jg@mail.gmail.com>
In-Reply-To: <CANzuiCj7b8iX55xyPtjJAdi6bKKCiHM4g5LL3-3jorm7m-b8jg@mail.gmail.com>
From: Phillipp Schoppmann <schoppmann@google.com>
Date: Sat, 8 Jan 2022 13:13:05 +0100
Message-ID: <CAFO2ENkzqkjWYfsgkb9GzP+ER63=OrKo7MUoH-yUuhHn=aZ5KA@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Roman Danyliw <rdd@cert.org>, "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="00000000000043cb0105d51107d5"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/_IrfmSFvFzIFkWGz51_SS4PXaYs>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sat, 08 Jan 2022 12:13:28 -0000

--00000000000043cb0105d51107d5
Content-Type: text/plain; charset="UTF-8"

   1. Yes
   2. Yes
   3. Yes
   4. Yes


On Fri, Jan 7, 2022 at 5:31 PM Erik Taubeneck <erik.taubeneck@gmail.com>
wrote:

> (1) Yes
> (2) Yes
> (3) Yes
> (4) Possibly, pending commitment and resourcing.
>
> Erik Taubeneck
>
> On Fri, Jan 7, 2022 at 8:02 AM Christopher Patton <cpatton=
> 40cloudflare.com@dmarc.ietf.org> wrote:
>
>> (1) Yes
>> (2) Yes
>> (3) Yes
>> (4) Yes
>>
>> Chris P.
>>
>> On Wed, Jan 5, 2022 at 1:19 PM Roman Danyliw <rdd@cert.org> wrote:
>>
>>> Hi!
>>>
>>> Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF
>>> 112.  The meeting notes [1] captured four major themes of feedback based on
>>> the presented charter [2] (version 00-00).  Specifically:
>>>
>>> ==[ snip from meeting notes ]==
>>> (a) Generalize the text to make it clear that flexibility for
>>> alternatives and additions to approach is possible (i.e., more than Prio,
>>> heavy hitters)
>>> (b) Refine the definition of aggregations so as to not restrict
>>> alternatives
>>> (c) Add a work item(s) to document the abuse cases and associated
>>> mitigations from the perspective of all participants in the architecture
>>> (d) Revision of the WG name to convey a more narrow scope
>>> ==[ snip ]==
>>>
>>> In response to this feedback, the proposed charter text has been revised
>>> at [3] (version 00-01).  Also, see [4] for a diff of this revision against
>>> the original text presented at IETF 112.  For convenience, this charter
>>> text is loaded into the datatracker using the BOF name (PRIV). As there
>>> wasn't consensus on "PRIV" as the WG name per feedback (d) or the now
>>> proposed alternative of "PPM", it didn't make sense to ask the Secretariat
>>> to create a new entry in the datatracker yet.  If consensus is found, a new
>>> datatracker entry will be made prior to sending the charter to the IESG.
>>>
>>> This email is intended to start a thread to gauge interest in forming a
>>> WG around this new, refined charter language [3].  Please respond to the
>>> list:
>>>
>>> (1) Do you support the charter text [3]? Or do you have objections or
>>> blocking concerns (please describe what they might be)?
>>>
>>> If you do support the charter text:
>>> (2) Are you willing to author or participate in the developed of the WG
>>> drafts?
>>> (3) Are you willing to review the WG drafts?
>>> (4) Are you interested in implementing the WG drafts?
>>>
>>> This call for consensus will end on Thursday, January 20.
>>>
>>> Regards,
>>> Roman
>>>
>>> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
>>> [2] charter-ietf-priv-00-00 -
>>> https://datatracker.ietf.org/doc/charter-ietf-priv/00-00/
>>> [3] charter-ietf-priv-00-01 -
>>> https://datatracker.ietf.org/doc/charter-ietf-priv/00-01/
>>> [4] diff of 00-00 vs. 00-01 --
>>> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt
>>> [5] https://datatracker.ietf.org/wg/priv/about/
>>>
>>> --
>>> Ppm mailing list
>>> Ppm@ietf.org
>>> https://www.ietf.org/mailman/listinfo/ppm
>>>
>> --
>> Ppm mailing list
>> Ppm@ietf.org
>> https://www.ietf.org/mailman/listinfo/ppm
>>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--00000000000043cb0105d51107d5
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><ol><li>Yes</li><li>Yes</li><li>Yes</li><li>Yes</li></ol><=
/div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">O=
n Fri, Jan 7, 2022 at 5:31 PM Erik Taubeneck &lt;<a href=3D"mailto:erik.tau=
beneck@gmail.com">erik.taubeneck@gmail.com</a>&gt; wrote:<br></div><blockqu=
ote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px=
 solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">(1) Yes<div>(2) =
Yes</div><div>(3) Yes</div><div>(4) Possibly, pending commitment and resour=
cing.</div><div><br></div><div>Erik Taubeneck</div></div><br><div class=3D"=
gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Fri, Jan 7, 2022 at 8=
:02 AM Christopher Patton &lt;cpatton=3D<a href=3D"mailto:40cloudflare.com@=
dmarc.ietf.org" target=3D"_blank">40cloudflare.com@dmarc.ietf.org</a>&gt; w=
rote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0p=
x 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=
=3D"ltr"><div>(1) Yes<br></div><div>(2) Yes</div><div>(3) Yes</div><div>(4)=
 Yes</div><div><br></div><div>Chris P.</div></div><br><div class=3D"gmail_q=
uote"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 5, 2022 at 1:19 PM =
Roman Danyliw &lt;<a href=3D"mailto:rdd@cert.org" target=3D"_blank">rdd@cer=
t.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:=
1ex">Hi!<br>
<br>
Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF 11=
2.=C2=A0 The meeting notes [1] captured four major themes of feedback based=
 on the presented charter [2] (version 00-00).=C2=A0 Specifically:<br>
<br>
=3D=3D[ snip from meeting notes ]=3D=3D<br>
(a) Generalize the text to make it clear that flexibility for alternatives =
and additions to approach is possible (i.e., more than Prio, heavy hitters)=
<br>
(b) Refine the definition of aggregations so as to not restrict alternative=
s<br>
(c) Add a work item(s) to document the abuse cases and associated mitigatio=
ns from the perspective of all participants in the architecture<br>
(d) Revision of the WG name to convey a more narrow scope<br>
=3D=3D[ snip ]=3D=3D=C2=A0 <br>
<br>
In response to this feedback, the proposed charter text has been revised at=
 [3] (version 00-01).=C2=A0 Also, see [4] for a diff of this revision again=
st the original text presented at IETF 112.=C2=A0 For convenience, this cha=
rter text is loaded into the datatracker using the BOF name (PRIV). As ther=
e wasn&#39;t consensus on &quot;PRIV&quot; as the WG name per feedback (d) =
or the now proposed alternative of &quot;PPM&quot;, it didn&#39;t make sens=
e to ask the Secretariat to create a new entry in the datatracker yet.=C2=
=A0 If consensus is found, a new datatracker entry will be made prior to se=
nding the charter to the IESG.<br>
<br>
This email is intended to start a thread to gauge interest in forming a WG =
around this new, refined charter language [3].=C2=A0 Please respond to the =
list:<br>
<br>
(1) Do you support the charter text [3]? Or do you have objections or block=
ing concerns (please describe what they might be)?<br>
<br>
If you do support the charter text:<br>
(2) Are you willing to author or participate in the developed of the WG dra=
fts?<br>
(3) Are you willing to review the WG drafts?<br>
(4) Are you interested in implementing the WG drafts?<br>
<br>
This call for consensus will end on Thursday, January 20.<br>
<br>
Regards,<br>
Roman <br>
<br>
[1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=3D"n=
oreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-112-p=
riv/</a><br>
[2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-00/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-00/</a><br>
[3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/doc/c=
harter-ietf-priv/00-01/" rel=3D"noreferrer" target=3D"_blank">https://datat=
racker.ietf.org/doc/charter-ietf-priv/00-01/</a><br>
[4] diff of 00-00 vs. 00-01 -- <a href=3D"https://www.ietf.org/rfcdiff?url1=
=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmiles=
tones-00-00.txt&amp;url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fchart=
er-ietf-priv%2Fwithmilestones-00-01.txt" rel=3D"noreferrer" target=3D"_blan=
k">https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2=
Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-00.txt&amp;url2=3Dhttps%3A%2F%=
2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-priv%2Fwithmilestones-00-01.txt=
</a><br>
[5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"noreferr=
er" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--00000000000043cb0105d51107d5--


From nobody Sun Jan  9 00:16:31 2022
Return-Path: <do_not_reply@mnot.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 09AA83A0CD0 for <ppm@ietfa.amsl.com>; Sun,  9 Jan 2022 00:16:15 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=mnot.net header.b=mynO7Xb/; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=moB5wgaT
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id uHExkAaI_HBe for <ppm@ietfa.amsl.com>; Sun,  9 Jan 2022 00:16:09 -0800 (PST)
Received: from wout2-smtp.messagingengine.com (wout2-smtp.messagingengine.com [64.147.123.25]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 62D2E3A0CD8 for <ppm@ietf.org>; Sun,  9 Jan 2022 00:16:09 -0800 (PST)
Received: from compute3.internal (compute3.nyi.internal [10.202.2.43]) by mailout.west.internal (Postfix) with ESMTP id D45753200E79 for <ppm@ietf.org>; Sun,  9 Jan 2022 02:40:24 -0500 (EST)
Received: from mailfrontend1 ([10.202.2.162]) by compute3.internal (MEProxy); Sun, 09 Jan 2022 02:40:25 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mnot.net; h= content-type:mime-version:from:to:subject; s=fm1; bh=j03HAJTd0ZB o0Nav/65FJ2/bJiyWAeeRsddkaS9feQs=; b=mynO7Xb/c8K18fN39Xz10n0VCmK IDzyy4+rz/fV5ojsfSMtu8duCtepCQYm7QAAp+ExwLtfHCpWe8BbLHksRa1TRc7G 4Ul6rXH/vBMVMXUvVjuVgG+UiWqDHHJxyjDoESk3npR8nvzXlyhAQ7JbH/U1WQKG kwqzyeaJh/Dmg+F5fa+kXrvpu/bHbWIddvU1Lmtc7lGXxKVuayn5XZ0EEodN9qlH woGEWiYicCCUcKQk8oRQUIBiybYOHACJxsjvSGyrERsVHuIK+Syd4bMLOhyaoPRi xFEmu1OsWPehClB6GqGEWDx313oYgerSmxqDinQWWE94pA+3AtMx1xzWdyQ==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=content-type:from:mime-version:subject:to :x-me-proxy:x-me-proxy:x-me-sender:x-me-sender:x-sasl-enc; s= fm1; bh=j03HAJTd0ZBo0Nav/65FJ2/bJiyWAeeRsddkaS9feQs=; b=moB5wgaT 4S2WFM9gF5oFjTJH6QmKAHSvmhZqCvkcm6K3WsrwxqN8ECmX0PuSv8StsfgXpuIy 6A2xYpfteTA1UrJyHqep436rwJ9CoXwTrie5gw4pVdXJGXqdt/2MdCT6DD1ibd5T zqEjjVIJ4ImCQ0/FIRAGC2tsmMWDeUZtKsOZICXXGbV4b05J+UffLgfsX8IqUXSV hqe6wsUIlconXvxiXTfL9bVEgELyf0Rf1kgNo3vVMg3a1ssZihqFYS5R93qe1vcz 0Z4Op/pXEzi4yqta3Z1efMm/Op8kBAg4QgU/UZvKWIQgjj1126hK+2K0gEjz/sW+ kag5+03EwAF1oA==
X-ME-Sender: <xms:aJHaYc42n-UUXNkj7c8sMsmhskgZPv47yuPzkOi7Gjnszd7-ySrBaQ> <xme:aJHaYd6F5UF2_F876dyZQH9Vnr1FZ_3Oq0pdVzRFS8ewdMaGOw-sdLJUTeADuWcqj roVZa8DP_G6pmT8uA>
X-ME-Received: <xmr:aJHaYbfLBylcFUZx5KxYYN9WGJpKVcwFkpdxvEpOf-2UPc12_t6DBHoIk4yH8Hxc-gkB4biGHGSCEIed6aG5ILad2nZdzRohGt80ztL_w_mBe1Kx3AmYpwhKrKO3X08yEMjF>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvuddrudegjedgtdejucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucfpohcuuggrthgvuchfihgvlhguucdlgeelmdenuc fjughrpegtggfhvffusegrtddtredttdejnecuhfhrohhmpeftvghpohhsihhtohhrhicu tegtthhivhhithihucfuuhhmmhgrrhihuceuohhtuceoughopghnohhtpghrvghplhihse hmnhhothdrnhgvtheqnecuggftrfgrthhtvghrnhepkeefvdduteejvdefkeehieevuefg fefhteetveegffekffefteffvdelheduieetnecuffhomhgrihhnpehgihhthhhusgdrtg homhenucevlhhushhtvghrufhiiigvpeehnecurfgrrhgrmhepmhgrihhlfhhrohhmpegu ohgpnhhothgprhgvphhlhiesmhhnohhtrdhnvght
X-ME-Proxy: <xmx:aJHaYRKT7zwIeUSm9p9IIlK3haix7kyEPXCvjriJJIrTTCcU6iGmIg> <xmx:aJHaYQKP_5zpnaKRrSO1fySb9LGfQh5M4rKkXJmufh8jyAyl9vB-KA> <xmx:aJHaYSxMCobg74u2DCqibbB9fo_Kgx_QmXh_pwhH9rGg9_bBJGjmLA> <xmx:aJHaYXW62JwM0R_P9HWyN8FC_diLv1DtTpw5NZJisOqC9iSVOyFppQ>
Received: by mail.messagingengine.com (Postfix) with ESMTPA for <ppm@ietf.org>;  Sun, 9 Jan 2022 02:40:24 -0500 (EST)
Content-Type: multipart/alternative; boundary="===============0622316700010533095=="
MIME-Version: 1.0
From: Repository Activity Summary Bot <do_not_reply@mnot.net>
To: ppm@ietf.org
Message-Id: <20220109081609.62D2E3A0CD8@ietfa.amsl.com>
Date: Sun,  9 Jan 2022 00:16:09 -0800 (PST)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/peNbkFT_gUAxS48tpc77Ua5DLt8>
Subject: [Ppm] Weekly github digest (Privacy Preserving Measurement Activity Summary)
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 09 Jan 2022 08:16:15 -0000

--===============0622316700010533095==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"; format="flowed"




Events=20

Issues
------
* abetterinternet/ppm-specification (+1/-1/=F0=9F=92=AC7)
  1 issues created:
  - "Heavy Hitters" is now "Poplar" (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/issues/176=20

  3 issues received 7 new comments:
  - #173 addressing collusion (1 by npdoty)
    https://github.com/abetterinternet/ppm-specification/issues/173=20
  - #161 Rotation of shared secrets in PPM task parameters (1 by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/161=20
  - #130 Cost implications of single leader model (5 by cjpatton, eriktaube=
neck)
    https://github.com/abetterinternet/ppm-specification/issues/130=20

  1 issues closed:
  - "Heavy Hitters" is now "Poplar" https://github.com/abetterinternet/ppm-=
specification/issues/176=20



Pull requests
-------------
* abetterinternet/ppm-specification (+2/-2/=F0=9F=92=AC3)
  2 pull requests submitted:
  - Poplar, with one 'o' (by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/pull/178=20
  - Rename "Heavy Hitters" to "Poplar" (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/177=20

  1 pull requests received 3 new comments:
  - #177 Rename "Heavy Hitters" to "Poplar" (3 by cjpatton, ekr)
    https://github.com/abetterinternet/ppm-specification/pull/177=20

  2 pull requests merged:
  - Poplar, with one 'o'
    https://github.com/abetterinternet/ppm-specification/pull/178=20
  - Rename "Heavy Hitters" to "Poplar"
    https://github.com/abetterinternet/ppm-specification/pull/177=20


Repositories tracked by this digest:
-----------------------------------
* https://github.com/abetterinternet/ppm-specification

--===============0622316700010533095==
Content-Type: text/html; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html lang=3D"en">
<head>
<meta charset=3D"utf-8">
<title>Weekly github digest (Privacy Preserving Measurement Activity Summar=
y)</title>
<style>
body { font-family: Gotham, "Helvetica Neue", Helvetica, Arial, sans-serif;=
 font-size: 14px; }
h2 { margin-top: 3em; color: #A52A2A; font-style: italic; font-weight: norm=
al; }
h3 { margin-bottom:0; margin-top: 2em; font-size: 1.2em; }
h1+h2 { margin-top: 1em; }
a { color: #bb6219; text-decoration: none; }
li { margin-bottom: .35em; }
.repos { margin-bottom: 0; margin-top:0; line-height: 1.2; }
.new { color: red; }
.label { display: inline;
	padding: .2em .6em .3em;
	font-size: 75%;
	font-weight: 700;
	line-height: 1;
	color: #fff;
	text-align: center;
	white-space: nowrap;
	vertical-align: baseline;
	border-radius: .25em;
}
details {
	margin-top: 8em;
	}
summary {
	margin-bottom: 1em;
	cursor: pointer;
}
</style>
</head>

<body>
<h1>Sunday January 09, 2022</h1>

<p>Events </p>

<h2>Issues</h2>

<h3>abetterinternet/ppm-specification (+1/-1/=F0=9F=92=AC7)</h3>
  <p class=3D"new">1 issues created:</p>
  <ul>
  <li>#176 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/176">&quot;Heavy Hitters&quot; is now &quot;Poplar&quot;</a> (by cjp=
atton) </li>
  </ul>

  <p>3 issues received 7 new comments:</p>
  <ul>
  <li>#173 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/173">addressing collusion</a> (1 by npdoty) </li>
 =20
  <li>#161 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/161">Rotation of shared secrets in PPM task parameters</a> (1 by tge=
oghegan) </li>
 =20
  <li>#130 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/130">Cost implications of single leader model</a> (5 by cjpatton, er=
iktaubeneck) </li>
  </ul>

  <p>1 issues closed:</p>
  <ul>
  <li>#176 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/176">&quot;Heavy Hitters&quot; is now &quot;Poplar&quot;</a> </li>
  </ul>



<h2>Pull requests</h2>
<h3>abetterinternet/ppm-specification (+2/-2/=F0=9F=92=AC3)</h3>
  <p class=3D"new">2 pull requests submitted:</p>
  <ul>
  <li>#178 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/178">Poplar, with one &#x27;o&#x27;</a> (by tgeoghegan) </li>
 =20
  <li>#177 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/177">Rename &quot;Heavy Hitters&quot; to &quot;Poplar&quot;</a> (by cj=
patton) </li>
  </ul>

  <p>1 pull requests received 3 new comments:</p>
  <ul>
  <li>#177 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/177">Rename &quot;Heavy Hitters&quot; to &quot;Poplar&quot;</a> (3 by =
cjpatton, ekr) </li>
  </ul>

  <p>2 pull requests merged:</p>
  <ul>
  <li>#178 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/178">Poplar, with one &#x27;o&#x27;</a> </li>
 =20
  <li>#177 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/177">Rename &quot;Heavy Hitters&quot; to &quot;Poplar&quot;</a> </li>
  </ul>


  <details>
    <summary>Repositories tracked by this digest:</summary>
<ul class=3D"repos">
  <li><a href=3D"https://github.com/abetterinternet/ppm-specification">http=
s://github.com/abetterinternet/ppm-specification</a></li>
</ul>
</details>
</body>
</html>

--===============0622316700010533095==--


From nobody Wed Jan 12 09:54:00 2022
Return-Path: <rdd@cert.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 9B1EC3A16AF for <ppm@ietfa.amsl.com>; Wed, 12 Jan 2022 09:53:58 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.899
X-Spam-Level: 
X-Spam-Status: No, score=-1.899 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=seicmu.onmicrosoft.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id EUfO0JknvB6l for <ppm@ietfa.amsl.com>; Wed, 12 Jan 2022 09:53:54 -0800 (PST)
Received: from USG02-CY1-obe.outbound.protection.office365.us (mail-cy1usg02on071b.outbound.protection.office365.us [IPv6:2001:489a:2202:d::71b]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 2385F3A16AD for <ppm@ietf.org>; Wed, 12 Jan 2022 09:53:53 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector5401; d=microsoft.com; cv=none; b=t4vgPelrThTPSIo35/LB6Wwm5Ei5xfHjxUxDGSnwGmctvEBn0Fb5lkOqcv5+b6DfaWJp5CsyEvPihPlZ6VO29Y0WOPpSixCidDUPcv7wgmGZUZ6RtfWNPIRBOz4JvwaB702LkLmIHqyDY1YkcLK/cpas05r+zsdKeVVxIHRIN1secJ3dj13y/NHVS3xx04QgvHWlVwFLqQJJcPwTOPYrOu3nt/+GCd+t2OpdfSaRNmnCLPNAfVGU/O4PtFTv9V1HobjgW0DBs9zw21kczptUs8r6xEOVWCIslcYSNzBcsKlFg+s3M9NKLms5QJmIrS38+3/fUIo+R3ojsJY6mBpamg==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector5401; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=IpUKeY46m26m9eyqTZ1D7x0hNs9iqYvtI0aSLpoCzRc=; b=gtytOVCnePP2G8YSDHfWavvpxbDMe4HI+AN3j+enaX1eSioXrejXDWnV+2v6veQCTva9rkjrLr0uMkWb32VABVDXkX595wedVHhb2/NBiZBRcZWVgf8lyFDlME/MIOG1Dt5i5JzK5TRe7O5/H0XwQcntfz5N+e44sKXJXyd9ex+jI1rgheqaJ8OruNgm7+ifgoqKGc64boWp0AVTzxKR2Y/Xc7Rf3EUbX0oWn6n5NgYayYB4hPGLezVE1GvFqjSwcxoQEU5OGXzv8G/Fx+IDkYibPK2v/c9GYpJ5ARuHJjur3yWmajTKViu3jwGj17VVifpfSnFmbJ3b6jtOjYhe1Q==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cert.org; dmarc=pass action=none header.from=cert.org; dkim=pass header.d=cert.org; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=seicmu.onmicrosoft.com; s=selector1-seicmu-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=IpUKeY46m26m9eyqTZ1D7x0hNs9iqYvtI0aSLpoCzRc=; b=hJFT0HnUdQ09ObseY6GlufIqcP4Y8D0PGSlICy7Hdxnz9AhCds38teIQxaYfq2+V62LYrgbFdv12dWsqVGNzFpgpZFd5UcbMLB+zgySPsf3/rZbYb8Bq3j0M5cC1Un8w6XwNB6+kglySt9/cvfBa6nG4VBI0Aoy4QPuIHGt7uVU=
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::11) by BN2P110MB0961.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:16b::9) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4888.9; Wed, 12 Jan 2022 17:53:43 +0000
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::f0be:6d5:6544:cce0]) by BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::f0be:6d5:6544:cce0%4]) with mapi id 15.20.4888.010; Wed, 12 Jan 2022 17:53:43 +0000
From: Roman Danyliw <rdd@cert.org>
To: "ppm@ietf.org" <ppm@ietf.org>
Thread-Topic: Call for consensus on updated charter
Thread-Index: AdgCd8GCV7y9nJtBRKqSR9LxS59tQQFZVVmg
Date: Wed, 12 Jan 2022 17:53:43 +0000
Message-ID: <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cert.org;
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 70ad461f-6fff-49bc-6a90-08d9d5f47993
x-ms-traffictypediagnostic: BN2P110MB0961:EE_
x-microsoft-antispam-prvs: <BN2P110MB0961F09AB76400D51B5ED3C1DC529@BN2P110MB0961.NAMP110.PROD.OUTLOOK.COM>
x-ms-oob-tlc-oobclassifiers: OLM:10000;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM; PTR:; CAT:NONE;  SFS:(366004)(86362001)(82960400001)(33656002)(8676002)(55016003)(8936002)(15650500001)(9686003)(83380400001)(5660300002)(52536014)(2906002)(66476007)(66556008)(64756008)(498600001)(76116006)(6916009)(186003)(6506007)(7696005)(53546011)(26005)(66946007)(38100700002)(66446008)(38070700005)(122000001)(71200400001)(966005); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: =?us-ascii?Q?cYIWoBOW35GneESABgjGq1S99aryPWwvoYN1PIPuyNjfT1jA1o3eMoofmoG4?= =?us-ascii?Q?IpKy+aZh/tGXaY2Ao2uDM0PpKLyKZp1LL9D6RQteHIJlwUAfCnLixfVnnZsr?= =?us-ascii?Q?4v6uq2aud42ETpI3OOJj6E/MxL+xqlyDYP4J5f8z8FBrlcScRBJq8CT0ZqN1?= =?us-ascii?Q?W4trfVu9vajXeuY1jhIfzhmNBac3IqU9L2g/IW96gLupqN/xVfkHrsF0Nt7S?= =?us-ascii?Q?A7w8yYkYTqpDCWo5ksAyCWfO4qmefOyBtsEqmAeWvFCok9jQzekG5Y2e9BjG?= =?us-ascii?Q?Io+0BxBjbhYSeTYGSVIv2Hv95eriTwfYq38QDaFhpK/5mwYgJ5cp1/uG5vke?= =?us-ascii?Q?wfcPUB7YInWPIik1I/SB7DqCIYCye1tXWJ9xpGcCBNdVIG+QPwpxEp3mWpcU?= =?us-ascii?Q?n97LlIPec456noGqDY2KQl0KUtjauPvPpYMmKKcmJCRawp2CZs5Qx/Ncsrru?= =?us-ascii?Q?hYTcf0gkUX9rPyvjw9FFgXcERfR3/blcuGg/oDVHO7lI8qv9/14+VPQMtmqH?= =?us-ascii?Q?0Yhn1D+Bt/J6Q6TKACyS586DVQjbLIF77l+jGFdvHA7hV3uontzwhfLLnaUY?= =?us-ascii?Q?Z27KDpaqnijB+/hqXTTa9Yd5PbBu2EwqdQd5PQnqeQqWwvji1oGiA2BWSYxK?= =?us-ascii?Q?KeMarnEFuPvXKi+PSDpfWX330XL7K61nq+ATJqQ3WQYjCFAJGsoK2zS4Quq9?= =?us-ascii?Q?SbLUeufGH0/rMeq1+r/W8VaRSb0L4OMng82AkLLSScVZy1BXV6L59fkFw0Ji?= =?us-ascii?Q?bkU5B6COHx7zWy6nj46i8If6l8qkTH/CjvCv+ctHLUs7YxtnqSLpggs5aRA4?= =?us-ascii?Q?XzV0zD6osZA3pAF9JrTD3xRPlKU7KIos6JgGfHkMKIdaU0ftmtaWNgZvlkCK?= =?us-ascii?Q?yaREm4/KOaIYP0HfhoofYSBFFY+J0NhcGEL5wTTUsU+9y2AGncNFDsPw2amX?= =?us-ascii?Q?bD7ZUh0A5RTpISvf3n4PbnyvpsSJrVBmjy8ug2K87pVJoxoLfLe2XlAEgEhQ?= =?us-ascii?Q?DfcVFmkUb50oBzJsOweadTa5eNigMUrjzUTQZkOQ5Bwc2ppM+tg8kgSkN1+W?= =?us-ascii?Q?k95TmM2LOIMDU41m+KcVaTwWe281wn/rGZ7RPgAjWPQv/sZEaJcPjV3Cr0Nz?= =?us-ascii?Q?qmYWIGDHDMh8lRAqCndnCcq0AWgp2HrzEYciXWobYT8T/JkVFo+Rjkk5W7Fg?= =?us-ascii?Q?y6CqJQVC/uKINlRxGYbGzYWrxGkX+bLVKinyy2rK45R/+g8dtDyomcQUFyU?= =?us-ascii?Q?=3D?=
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: cert.org
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM
X-MS-Exchange-CrossTenant-Network-Message-Id: 70ad461f-6fff-49bc-6a90-08d9d5f47993
X-MS-Exchange-CrossTenant-originalarrivaltime: 12 Jan 2022 17:53:43.7250 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 95a9dce2-04f2-4043-995d-1ec3861911c6
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN2P110MB0961
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/yIi-QAW29d4ZmJ0EnBuxZYQQKbg>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 12 Jan 2022 17:53:59 -0000

Hi!

Thanks to all that have responded to this email.

A reminder that additional input is welcome and would be appreciated by Thu=
rsday, January 20.

Regards,
Roman

> -----Original Message-----
> From: Ppm <ppm-bounces@ietf.org> On Behalf Of Roman Danyliw
> Sent: Wednesday, January 5, 2022 4:19 PM
> To: ppm@ietf.org
> Subject: [Ppm] Call for consensus on updated charter
>=20
> Hi!
>=20
> Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF =
112.
> The meeting notes [1] captured four major themes of feedback based on the
> presented charter [2] (version 00-00).  Specifically:
>=20
> =3D=3D[ snip from meeting notes ]=3D=3D
> (a) Generalize the text to make it clear that flexibility for alternative=
s and
> additions to approach is possible (i.e., more than Prio, heavy hitters)
> (b) Refine the definition of aggregations so as to not restrict alternati=
ves
> (c) Add a work item(s) to document the abuse cases and associated mitigat=
ions
> from the perspective of all participants in the architecture
> (d) Revision of the WG name to convey a more narrow scope =3D=3D[ snip ]=
=3D=3D
>=20
> In response to this feedback, the proposed charter text has been revised =
at [3]
> (version 00-01).  Also, see [4] for a diff of this revision against the o=
riginal text
> presented at IETF 112.  For convenience, this charter text is loaded into=
 the
> datatracker using the BOF name (PRIV). As there wasn't consensus on "PRIV=
" as
> the WG name per feedback (d) or the now proposed alternative of "PPM", it
> didn't make sense to ask the Secretariat to create a new entry in the
> datatracker yet.  If consensus is found, a new datatracker entry will be =
made
> prior to sending the charter to the IESG.
>=20
> This email is intended to start a thread to gauge interest in forming a W=
G
> around this new, refined charter language [3].  Please respond to the lis=
t:
>=20
> (1) Do you support the charter text [3]? Or do you have objections or blo=
cking
> concerns (please describe what they might be)?
>=20
> If you do support the charter text:
> (2) Are you willing to author or participate in the developed of the WG d=
rafts?
> (3) Are you willing to review the WG drafts?
> (4) Are you interested in implementing the WG drafts?
>=20
> This call for consensus will end on Thursday, January 20.
>=20
> Regards,
> Roman
>=20
> [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> [2] charter-ietf-priv-00-00 - https://datatracker.ietf.org/doc/charter-ie=
tf-
> priv/00-00/
> [3] charter-ietf-priv-00-01 - https://datatracker.ietf.org/doc/charter-ie=
tf-
> priv/00-01/
> [4] diff of 00-00 vs. 00-01 --
> https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2F=
doc
> %2Fcharter-ietf-priv%2Fwithmilestones-00-
> 00.txt&url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-
> priv%2Fwithmilestones-00-01.txt
> [5] https://datatracker.ietf.org/wg/priv/about/
>=20
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm


From nobody Wed Jan 12 14:42:59 2022
Return-Path: <csharrison@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 602DC3A0139 for <ppm@ietfa.amsl.com>; Wed, 12 Jan 2022 14:42:57 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 9GA9rNJy10cU for <ppm@ietfa.amsl.com>; Wed, 12 Jan 2022 14:42:52 -0800 (PST)
Received: from mail-ed1-x52e.google.com (mail-ed1-x52e.google.com [IPv6:2a00:1450:4864:20::52e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 2B4863A0147 for <ppm@ietf.org>; Wed, 12 Jan 2022 14:42:52 -0800 (PST)
Received: by mail-ed1-x52e.google.com with SMTP id o6so15874994edc.4 for <ppm@ietf.org>; Wed, 12 Jan 2022 14:42:52 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=Y7HiA59lRJv4ieA8o6/CfN6ASd8ISnplME0sOH/Gxs0=; b=CyRWzJpOG/Lt5zFhGLwZXZAOOjjeujOZQEQlqYaVezYOHOy20Gzij1bB5l5AaHj2wS 6+7+fn/rGIiuQ+JIyRfY2XrCCLGN20UTn2dhOaZuRsIGUbFLEKRGLwfr5NiHs+yBbte5 FLvPhSw4D57z6wb7NcAFpdjUIjPPUcvC+sxpM2SkTYkHf62zpNo14SEilNwQTLpC15Ph VuIDG0VtoBTC9EO7eqIqKUd1SPgj2FOI4Ce6ZrleTPsoCA/ykDghnS15sBf5P/OuJgmO Ju/JDsRXshdCIeo3MnfGfeifaoeV50TeSFycIoRiZQsv7UnHWmlWRnoCW+u1x4lvsVg9 0Kog==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=Y7HiA59lRJv4ieA8o6/CfN6ASd8ISnplME0sOH/Gxs0=; b=uZHAZvYM3F2+7e8et/FhHTFZUEU9QbUAL1eNvMlip+EoyWjeruMEZTBxlCEQw8myad v1w7V6dZleQoY5EHsHZaXZPDg1LwiVfqcZ64PUdXHGiaEb0qswgjyFEiojfxHf5eqNFi dnS5Abhxd2QWWK8av3pO3l4pgbqbPFrgZCWmFyY+Sqzjq/MHfvLU9fHVBKe7UEXOEyZd nski4j+bzBxQm92r0vLCmAHY6VRK/OcFnLHsiSuuN+O28h44xl2AycyIA9gNRv+pGvy9 YHzhVLq8viKF3udhKCcoI/o/V0kF3FrLX0Px+OAA5R8UAdlxxOj6U2+rcIO2J8e1em1l MJpw==
X-Gm-Message-State: AOAM533FkVN5o1S+5EDBGPBPb3C6iKXjSCG/ybnVlz6xbSiV5GsujfIR aZ5Vhh52Lf57jOZ1Jj4DsYL8Oce3Pck6WF0UE8AkpOzgGC4=
X-Google-Smtp-Source: ABdhPJxz/h7rTiQLFByaQjLSpC+wQOGgwSoXj7GrYrU5ueiBRQPYqkaXKxVcebxreHHIyqOz7rPS5/cmMhjcVKBl9+c=
X-Received: by 2002:aa7:d74f:: with SMTP id a15mr1661912eds.163.1642027369802;  Wed, 12 Jan 2022 14:42:49 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Charlie Harrison <csharrison@google.com>
Date: Wed, 12 Jan 2022 16:42:38 -0600
Message-ID: <CADjAqN6j+hja8mdmQduWpE=L6a1rzh0C6Nv74gvi_LpzEqv4rQ@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="00000000000016f21005d56a4ada"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/gn2vzSEvirQAnQp8Jatu2m4azs0>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 12 Jan 2022 22:42:57 -0000

--00000000000016f21005d56a4ada
Content-Type: text/plain; charset="UTF-8"

1. Yes
2. Yes
3. Yes
4. Possibly

On Wed, Jan 12, 2022 at 11:54 AM Roman Danyliw <rdd@cert.org> wrote:

> Hi!
>
> Thanks to all that have responded to this email.
>
> A reminder that additional input is welcome and would be appreciated by
> Thursday, January 20.
>
> Regards,
> Roman
>
> > -----Original Message-----
> > From: Ppm <ppm-bounces@ietf.org> On Behalf Of Roman Danyliw
> > Sent: Wednesday, January 5, 2022 4:19 PM
> > To: ppm@ietf.org
> > Subject: [Ppm] Call for consensus on updated charter
> >
> > Hi!
> >
> > Happy New Year! Thanks for the robust discussion at the PRIV BOF at IETF
> 112.
> > The meeting notes [1] captured four major themes of feedback based on the
> > presented charter [2] (version 00-00).  Specifically:
> >
> > ==[ snip from meeting notes ]==
> > (a) Generalize the text to make it clear that flexibility for
> alternatives and
> > additions to approach is possible (i.e., more than Prio, heavy hitters)
> > (b) Refine the definition of aggregations so as to not restrict
> alternatives
> > (c) Add a work item(s) to document the abuse cases and associated
> mitigations
> > from the perspective of all participants in the architecture
> > (d) Revision of the WG name to convey a more narrow scope ==[ snip ]==
> >
> > In response to this feedback, the proposed charter text has been revised
> at [3]
> > (version 00-01).  Also, see [4] for a diff of this revision against the
> original text
> > presented at IETF 112.  For convenience, this charter text is loaded
> into the
> > datatracker using the BOF name (PRIV). As there wasn't consensus on
> "PRIV" as
> > the WG name per feedback (d) or the now proposed alternative of "PPM", it
> > didn't make sense to ask the Secretariat to create a new entry in the
> > datatracker yet.  If consensus is found, a new datatracker entry will be
> made
> > prior to sending the charter to the IESG.
> >
> > This email is intended to start a thread to gauge interest in forming a
> WG
> > around this new, refined charter language [3].  Please respond to the
> list:
> >
> > (1) Do you support the charter text [3]? Or do you have objections or
> blocking
> > concerns (please describe what they might be)?
> >
> > If you do support the charter text:
> > (2) Are you willing to author or participate in the developed of the WG
> drafts?
> > (3) Are you willing to review the WG drafts?
> > (4) Are you interested in implementing the WG drafts?
> >
> > This call for consensus will end on Thursday, January 20.
> >
> > Regards,
> > Roman
> >
> > [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> > [2] charter-ietf-priv-00-00 -
> https://datatracker.ietf.org/doc/charter-ietf-
> > priv/00-00/
> > [3] charter-ietf-priv-00-01 -
> https://datatracker.ietf.org/doc/charter-ietf-
> > priv/00-01/
> > [4] diff of 00-00 vs. 00-01 --
> >
> https://www.ietf.org/rfcdiff?url1=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc
> > %2Fcharter-ietf-priv%2Fwithmilestones-00-
> > 00.txt&url2=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-
> > priv%2Fwithmilestones-00-01.txt
> > [5] https://datatracker.ietf.org/wg/priv/about/
> >
> > --
> > Ppm mailing list
> > Ppm@ietf.org
> > https://www.ietf.org/mailman/listinfo/ppm
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--00000000000016f21005d56a4ada
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">1. Yes<div>2. Yes</div><div>3. Yes</div><div>4. Possibly</=
div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_at=
tr">On Wed, Jan 12, 2022 at 11:54 AM Roman Danyliw &lt;<a href=3D"mailto:rd=
d@cert.org">rdd@cert.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex">Hi!<br>
<br>
Thanks to all that have responded to this email.<br>
<br>
A reminder that additional input is welcome and would be appreciated by Thu=
rsday, January 20.<br>
<br>
Regards,<br>
Roman<br>
<br>
&gt; -----Original Message-----<br>
&gt; From: Ppm &lt;<a href=3D"mailto:ppm-bounces@ietf.org" target=3D"_blank=
">ppm-bounces@ietf.org</a>&gt; On Behalf Of Roman Danyliw<br>
&gt; Sent: Wednesday, January 5, 2022 4:19 PM<br>
&gt; To: <a href=3D"mailto:ppm@ietf.org" target=3D"_blank">ppm@ietf.org</a>=
<br>
&gt; Subject: [Ppm] Call for consensus on updated charter<br>
&gt; <br>
&gt; Hi!<br>
&gt; <br>
&gt; Happy New Year! Thanks for the robust discussion at the PRIV BOF at IE=
TF 112.<br>
&gt; The meeting notes [1] captured four major themes of feedback based on =
the<br>
&gt; presented charter [2] (version 00-00).=C2=A0 Specifically:<br>
&gt; <br>
&gt; =3D=3D[ snip from meeting notes ]=3D=3D<br>
&gt; (a) Generalize the text to make it clear that flexibility for alternat=
ives and<br>
&gt; additions to approach is possible (i.e., more than Prio, heavy hitters=
)<br>
&gt; (b) Refine the definition of aggregations so as to not restrict altern=
atives<br>
&gt; (c) Add a work item(s) to document the abuse cases and associated miti=
gations<br>
&gt; from the perspective of all participants in the architecture<br>
&gt; (d) Revision of the WG name to convey a more narrow scope =3D=3D[ snip=
 ]=3D=3D<br>
&gt; <br>
&gt; In response to this feedback, the proposed charter text has been revis=
ed at [3]<br>
&gt; (version 00-01).=C2=A0 Also, see [4] for a diff of this revision again=
st the original text<br>
&gt; presented at IETF 112.=C2=A0 For convenience, this charter text is loa=
ded into the<br>
&gt; datatracker using the BOF name (PRIV). As there wasn&#39;t consensus o=
n &quot;PRIV&quot; as<br>
&gt; the WG name per feedback (d) or the now proposed alternative of &quot;=
PPM&quot;, it<br>
&gt; didn&#39;t make sense to ask the Secretariat to create a new entry in =
the<br>
&gt; datatracker yet.=C2=A0 If consensus is found, a new datatracker entry =
will be made<br>
&gt; prior to sending the charter to the IESG.<br>
&gt; <br>
&gt; This email is intended to start a thread to gauge interest in forming =
a WG<br>
&gt; around this new, refined charter language [3].=C2=A0 Please respond to=
 the list:<br>
&gt; <br>
&gt; (1) Do you support the charter text [3]? Or do you have objections or =
blocking<br>
&gt; concerns (please describe what they might be)?<br>
&gt; <br>
&gt; If you do support the charter text:<br>
&gt; (2) Are you willing to author or participate in the developed of the W=
G drafts?<br>
&gt; (3) Are you willing to review the WG drafts?<br>
&gt; (4) Are you interested in implementing the WG drafts?<br>
&gt; <br>
&gt; This call for consensus will end on Thursday, January 20.<br>
&gt; <br>
&gt; Regards,<br>
&gt; Roman<br>
&gt; <br>
&gt; [1] <a href=3D"https://datatracker.ietf.org/doc/minutes-112-priv/" rel=
=3D"noreferrer" target=3D"_blank">https://datatracker.ietf.org/doc/minutes-=
112-priv/</a><br>
&gt; [2] charter-ietf-priv-00-00 - <a href=3D"https://datatracker.ietf.org/=
doc/charter-ietf-" rel=3D"noreferrer" target=3D"_blank">https://datatracker=
.ietf.org/doc/charter-ietf-</a><br>
&gt; priv/00-00/<br>
&gt; [3] charter-ietf-priv-00-01 - <a href=3D"https://datatracker.ietf.org/=
doc/charter-ietf-" rel=3D"noreferrer" target=3D"_blank">https://datatracker=
.ietf.org/doc/charter-ietf-</a><br>
&gt; priv/00-01/<br>
&gt; [4] diff of 00-00 vs. 00-01 --<br>
&gt; <a href=3D"https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatrack=
er.ietf.org%2Fdoc" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.or=
g/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc</a><br>
&gt; %2Fcharter-ietf-priv%2Fwithmilestones-00-<br>
&gt; 00.txt&amp;url2=3Dhttps%3A%2F%<a href=3D"http://2Fdatatracker.ietf.org=
" rel=3D"noreferrer" target=3D"_blank">2Fdatatracker.ietf.org</a>%2Fdoc%2Fc=
harter-ietf-<br>
&gt; priv%2Fwithmilestones-00-01.txt<br>
&gt; [5] <a href=3D"https://datatracker.ietf.org/wg/priv/about/" rel=3D"nor=
eferrer" target=3D"_blank">https://datatracker.ietf.org/wg/priv/about/</a><=
br>
&gt; <br>
&gt; --<br>
&gt; Ppm mailing list<br>
&gt; <a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
&gt; <a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferre=
r" target=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--00000000000016f21005d56a4ada--


From nobody Sun Jan 16 00:11:38 2022
Return-Path: <do_not_reply@mnot.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 2E3023A0D90 for <ppm@ietfa.amsl.com>; Sun, 16 Jan 2022 00:11:36 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=mnot.net header.b=fjGzzqS/; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=losh/QNd
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id PIWU5q3Ebg24 for <ppm@ietfa.amsl.com>; Sun, 16 Jan 2022 00:11:31 -0800 (PST)
Received: from wout2-smtp.messagingengine.com (wout2-smtp.messagingengine.com [64.147.123.25]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 6F55B3A0D8E for <ppm@ietf.org>; Sun, 16 Jan 2022 00:11:31 -0800 (PST)
Received: from compute3.internal (compute3.nyi.internal [10.202.2.43]) by mailout.west.internal (Postfix) with ESMTP id D19F73200F9C for <ppm@ietf.org>; Sun, 16 Jan 2022 02:40:54 -0500 (EST)
Received: from mailfrontend2 ([10.202.2.163]) by compute3.internal (MEProxy); Sun, 16 Jan 2022 02:40:54 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mnot.net; h= content-type:mime-version:from:to:subject; s=fm1; bh=qCgS8wXBfk0 jxW/FyIaoe67UUnWhEQqjFTRwq47G0hs=; b=fjGzzqS/aH1+LPftMA5KKianti3 kcacfXqlJ2T2s/Kry0Mq8NynBOB1mYJBivB3yIxTJPlTtNnqRVpkOIWzoCxBsUSR lyU6tEOMzlJLMpZgUFkAi7FOIIuktVa20LmAZPUfmzPUYgDgdJGAzTz7RhCbgmg1 qxtVp+STpB/V7XBsPey/zacUttXHkPfL3Dvs0MuKe+/WfE2gY+SgLr9Fio8zUA3O HgbrNfNmGk0WJc5arP6xdlqjVO8Okvcun0BuYhGphWUOf5p6KjG9R2rVWIFEcSPT GuUpTeYW/MojwPw6qHYvSarTRKUvkUL/QWjm69NdEsSLxHgLdTfu4mare7Q==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=content-type:from:mime-version:subject:to :x-me-proxy:x-me-proxy:x-me-sender:x-me-sender:x-sasl-enc; s= fm1; bh=qCgS8wXBfk0jxW/FyIaoe67UUnWhEQqjFTRwq47G0hs=; b=losh/QNd hTBcXvqeN68OQtOC6ZkbkjoecVbmzuHdxonAWtPZDCUnU+802ZGQpmio4LA/7D8P ggnIBx5KHWkNPcDO46WybHV5Sm7C1qd61xMoDdbo9unKOFSsXwRxx4uasFcelPEa UlfpUrXP9LqGIGGahxcIqY1lXU8C7sPE5KeehzUKSTp2VleyiRD/+moLE5SIZo8C l70yZkkgH8N9I0RdQFDo7rt4BjXrJYaTp3PiEU1c2Tbln3R+PIvgPZjYi92tWv0m V7CXRaBqubgR6/6ErwZ7JnZlfoZfnJcjeqBeoQuiOkynI8bNL6jAmGJ6CFhVTWQc 4SS4/PrPKlY9sw==
X-ME-Sender: <xms:BszjYQ2QOsVcC3gK__um-OR_VHVGqSgf2ZNoqzR-4-qMns5MES7bbw> <xme:BszjYbEfTfF_UB2onVplebKhYJmGnidAOnIKzhEbHdzCnPMeGI7CZsKZE115huiwS wRyRxcwWE8wDv3ZKg>
X-ME-Received: <xmr:BszjYY7c27TOP7XnHWCsN_u_uKVs0ZdkQh1A3xbtHZgFBvosyJV3fU2I3YEUynSGcAE3KK4IR8mXiKEI9rFVOr9SdMpUEwu4gkeTBGoqmT8h8_3zfeYglHJZP9zse50DpQ>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrtdekgddutdelucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucfpohcuuggrthgvuchfihgvlhguucdlgeelmdenuc fjughrpegtggfhvffusegrtddtredttdejnecuhfhrohhmpeftvghpohhsihhtohhrhicu tegtthhivhhithihucfuuhhmmhgrrhihuceuohhtuceoughopghnohhtpghrvghplhihse hmnhhothdrnhgvtheqnecuggftrfgrthhtvghrnhepkeefvdduteejvdefkeehieevuefg fefhteetveegffekffefteffvdelheduieetnecuffhomhgrihhnpehgihhthhhusgdrtg homhenucevlhhushhtvghrufhiiigvpeefnecurfgrrhgrmhepmhgrihhlfhhrohhmpegu ohgpnhhothgprhgvphhlhiesmhhnohhtrdhnvght
X-ME-Proxy: <xmx:BszjYZ3oj0Wf2x4_ABa8YDCO5EguO6CYFLX7NK3Dr3O04MBCHX5EVg> <xmx:BszjYTHdQrwekAD5SpeMpiHl7YG3PMtEY-fZIRP2ue-c2EKYQ9qMJQ> <xmx:BszjYS_QRCO1MA5Cm-xRgdc9Azlt02XnP3hXrnLn9HxkFS5FcNun8Q> <xmx:BszjYYT_3--ATdqcSInEzCLVy1SvIvy7bFchIpstATMWwzsGKKo4bQ>
Received: by mail.messagingengine.com (Postfix) with ESMTPA for <ppm@ietf.org>;  Sun, 16 Jan 2022 02:40:53 -0500 (EST)
Content-Type: multipart/alternative; boundary="===============2004180077941240894=="
MIME-Version: 1.0
From: Repository Activity Summary Bot <do_not_reply@mnot.net>
To: ppm@ietf.org
Message-Id: <20220116081131.6F55B3A0D8E@ietfa.amsl.com>
Date: Sun, 16 Jan 2022 00:11:31 -0800 (PST)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/_-Sy_bsOXv0bkqFOlJy-AVPiuLg>
Subject: [Ppm] Weekly github digest (Privacy Preserving Measurement Activity Summary)
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 16 Jan 2022 08:11:36 -0000

--===============2004180077941240894==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"; format="flowed"




Events=20

Issues
------
* abetterinternet/ppm-specification (+1/-0/=F0=9F=92=AC3)
  1 issues created:
  - Anti-replay requires aggregators to store nonce sets indefinitely (by c=
jpatton)
    https://github.com/abetterinternet/ppm-specification/issues/180=20

  1 issues received 3 new comments:
  - #161 Rotation of shared secrets in PPM task parameters (3 by cjpatton, =
henrycg)
    https://github.com/abetterinternet/ppm-specification/issues/161=20



Pull requests
-------------
* abetterinternet/ppm-specification (+1/-1/=F0=9F=92=AC11)
  1 pull requests submitted:
  - DO NOTE MERGE Proposed interop target (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/179=20

  1 pull requests received 11 new comments:
  - #169 Revise anti-replay mechanism (11 by chris-wood, cjpatton, ekr)
    https://github.com/abetterinternet/ppm-specification/pull/169=20

  1 pull requests merged:
  - Revise anti-replay mechanism
    https://github.com/abetterinternet/ppm-specification/pull/169=20


Repositories tracked by this digest:
-----------------------------------
* https://github.com/abetterinternet/ppm-specification

--===============2004180077941240894==
Content-Type: text/html; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html lang=3D"en">
<head>
<meta charset=3D"utf-8">
<title>Weekly github digest (Privacy Preserving Measurement Activity Summar=
y)</title>
<style>
body { font-family: Gotham, "Helvetica Neue", Helvetica, Arial, sans-serif;=
 font-size: 14px; }
h2 { margin-top: 3em; color: #A52A2A; font-style: italic; font-weight: norm=
al; }
h3 { margin-bottom:0; margin-top: 2em; font-size: 1.2em; }
h1+h2 { margin-top: 1em; }
a { color: #bb6219; text-decoration: none; }
li { margin-bottom: .35em; }
.repos { margin-bottom: 0; margin-top:0; line-height: 1.2; }
.new { color: red; }
.label { display: inline;
	padding: .2em .6em .3em;
	font-size: 75%;
	font-weight: 700;
	line-height: 1;
	color: #fff;
	text-align: center;
	white-space: nowrap;
	vertical-align: baseline;
	border-radius: .25em;
}
details {
	margin-top: 8em;
	}
summary {
	margin-bottom: 1em;
	cursor: pointer;
}
</style>
</head>

<body>
<h1>Sunday January 16, 2022</h1>

<p>Events </p>

<h2>Issues</h2>

<h3>abetterinternet/ppm-specification (+1/-0/=F0=9F=92=AC3)</h3>
  <p class=3D"new">1 issues created:</p>
  <ul>
  <li>#180 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/180">Anti-replay requires aggregators to store nonce sets indefinite=
ly</a> (by cjpatton) </li>
  </ul>

  <p>1 issues received 3 new comments:</p>
  <ul>
  <li>#161 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/161">Rotation of shared secrets in PPM task parameters</a> (3 by cjp=
atton, henrycg) </li>
  </ul>




<h2>Pull requests</h2>
<h3>abetterinternet/ppm-specification (+1/-1/=F0=9F=92=AC11)</h3>
  <p class=3D"new">1 pull requests submitted:</p>
  <ul>
  <li>#179 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/179">DO NOTE MERGE Proposed interop target</a> (by cjpatton) </li>
  </ul>

  <p>1 pull requests received 11 new comments:</p>
  <ul>
  <li>#169 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/169">Revise anti-replay mechanism</a> (11 by chris-wood, cjpatton, ekr=
) </li>
  </ul>

  <p>1 pull requests merged:</p>
  <ul>
  <li>#169 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/169">Revise anti-replay mechanism</a> </li>
  </ul>


  <details>
    <summary>Repositories tracked by this digest:</summary>
<ul class=3D"repos">
  <li><a href=3D"https://github.com/abetterinternet/ppm-specification">http=
s://github.com/abetterinternet/ppm-specification</a></li>
</ul>
</details>
</body>
</html>

--===============2004180077941240894==--


From nobody Tue Jan 18 13:52:53 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 7884E3A0DFB for <ppm@ietfa.amsl.com>; Tue, 18 Jan 2022 13:52:51 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id rKlEmZeh4NH3 for <ppm@ietfa.amsl.com>; Tue, 18 Jan 2022 13:52:46 -0800 (PST)
Received: from mail-pl1-x62a.google.com (mail-pl1-x62a.google.com [IPv6:2607:f8b0:4864:20::62a]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id D66733A0DF3 for <ppm@ietf.org>; Tue, 18 Jan 2022 13:52:46 -0800 (PST)
Received: by mail-pl1-x62a.google.com with SMTP id b3so205769plc.7 for <ppm@ietf.org>; Tue, 18 Jan 2022 13:52:46 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to; bh=Jo42lWgTDh+RwKNy7yChbAQmo2cHXxGHC5M/IH3SekU=; b=ugOtloUylTtIHwYaFzmL55RGl2UqPPFfJUfJKMfM6ViczbN+BlbjUwZmMVXaILQ4o6 1HPTXOUCjQgLXpoiF5TFwSkSwC3Mnx/abT5TdOwuR4EtORCewQpyiQkRlwoj/0ixpgFb MvUl6D9Vpt+b5W2MGQajxWQdo99ha5xodOxUE=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to; bh=Jo42lWgTDh+RwKNy7yChbAQmo2cHXxGHC5M/IH3SekU=; b=rZWg91UtoFsTcw8LC7UKgwBYFeLPWJrZ3zT7qgbAiDETpS17QSnZXG+rXymVH6JPhU 1SbFjwJKGeiS+zAvyZ9flEC2MT8UMEe3Dd9zdg9WIwlzizYTjk4wDR3wbaer730FAxwP y64JQyqYYfJspH0coIUEri7/XkP7e2iS6189MUZ7OiZndQXFvdMYwHPS5V/F5dHj1QcD whvUqqG20hdbcviN/ulqceJI+yQ9RPL0zcw8WyvZccikkIRU4oZhGEVrNFOrORiPy6F4 wQz6/4twptGe5H6gdhp5KXPaskpNtMjPlRmZpRtkMoSIMMIYPTzCIVfcjPFjfJFa87jb JZng==
X-Gm-Message-State: AOAM530ap3dGcZKdtUCyIm5IhYVUVQXq2rrjguJcFB9y6wI+1Z4zc5os P+h2S6qcCQB//mNyvhLjmgoi9Y8397BdFA9neMkwO5opxXAyBg==
X-Google-Smtp-Source: ABdhPJz2GtPN5HKHboeBvQmyUXJ2gLqtAYiWkh/ox+ZcxtUVgP+OOHnGAvZU9QNE4m3fvJy9BRK6kgmO0NjuBwpor7U=
X-Received: by 2002:a17:90b:506:: with SMTP id r6mr659677pjz.128.1642542765051;  Tue, 18 Jan 2022 13:52:45 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi20oarjT6jBbtFtvQUMy5uRZTK6Z2YgBSE+wT33xQbkb+A@mail.gmail.com> <CAG2Zi21N3fHZpvXadLbt3rhC-5LzLVoPRZdj01ONJCj-4YsfdA@mail.gmail.com>
In-Reply-To: <CAG2Zi21N3fHZpvXadLbt3rhC-5LzLVoPRZdj01ONJCj-4YsfdA@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Tue, 18 Jan 2022 13:52:33 -0800
Message-ID: <CAG2Zi23diGTnq1_NW_KEke2P_NGJqDb9yCKOgMOw1TuW7B8U1g@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000000a41bf05d5e24af1"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/fMYi1QiuhlKz7Uy9w1NEF5HcZ1o>
Subject: Re: [Ppm] Revisit the operational requirements for the Helper?
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 18 Jan 2022 21:52:52 -0000

--0000000000000a41bf05d5e24af1
Content-Type: text/plain; charset="UTF-8"

Hi all, at the end of last week we merged the PR:
https://github.com/abetterinternet/ppm-specification/pull/169
An issue that we left open is how to prevent aggregators from having to
store nonce sets indefinitely. That issue is tracked here:
https://github.com/abetterinternet/ppm-specification/issues/180

Thanks to all who chime in on this.

Best,
Chris P.


On Mon, Jan 3, 2022 at 9:14 AM Christopher Patton <cpatton@cloudflare.com>
wrote:

> Happy new year friends! The PR with the suggested change to anti-replay is
> up to date and ready for another round of review:
> https://github.com/abetterinternet/ppm-specification/pull/169
>
> It would be great if, by the end of the week, we can get a sense of
> whether folks support (some version of) this change or not
>
> Cheers,
> Chris P.
>
> On Thu, Dec 2, 2021 at 4:04 PM Christopher Patton <cpatton@cloudflare.com>
> wrote:
>
>> Hi everyone,
>>
>> I've begun an implementation of the PPM spec with `prio3` [1]. One of the
>> design goals of the sepc was to make it as easy as possible for a Helper to
>> participate in the protocol, even at the expense of higher resource
>> requirements for the Leader. Based on my experience so far, I think we may
>> have over-indexed on this goal a bit. I propose that we revisit this and
>> consider balancing the work between the Aggregators more evenly. While at
>> it, we may be able to resolve [2].
>>
>> The spec requires that reports be processed in strictly ascending order,
>> where the order is defined in [3]. The ordering is based on a timestamp and
>> random number generated by the client and included in the report. I'll
>> refer to thus (time, rand) pair as the nonce. Ordering reports based on the
>> nonce allows the Helper to prevent replay attacks* without needing to store
>> the nonces. In fact, the only long-term storage it needs for anti-replay is
>> the most recently processed nonce. The downside is that the Leader has to
>> sort reports as they arrive, which makes horizontally scaling the Leader
>> more challenging.
>>
>> A simpler alternative that we had discussed previously is just having
>> each of the Aggregators remember the nonces of the reports they have
>> processed. This removes the Leader's coordination problem, at the cost of
>> additional storage requirements. In fact, both Aggregators need to store
>> the set of processed nonces for the duration of a task. This cost can be
>> mitigated by refusing to process reports that arrive, say, a day or two
>> late. (Of course, this would need to be a parameter that the Aggregators
>> agree on in advance.)
>>
>> Let's call these schemes option #1 and option #2 respectively. Previously
>> I was in favor of option #1, but after getting my hands a little dirty and
>> conferring with some of my colleagues, I've come to the conclusion that the
>> costs of #1 outweigh the costs of #2. I'm curious if others have thought
>> about this and whether they are strongly in favor of either option.
>>
>> Now, if we take option #2, we might also consider revisiting the question
>> raised in [2]. The spec currently requires the Clients to upload their
>> encrypted input shares to the Leader, who then piggy-backs the Helper's
>> input shares on its aggregate requests. Egress fees from cloud providers is
>> typically higher than ingress, which makes the status quo more expensive to
>> deploy than if the Clients uploaded their encrypted input shares directly
>> to each Aggregator.
>>
>> What might an alternative look like? The client would generate a nonce as
>> usual and upload the nonce and an encrypted input share to each of the
>> Aggregators. The aggregation flow would largely be the same: The Leader
>> sends an Aggregate Request to the Helper, which contains the nonce of the
>> report it wants to aggregate as usual. The difference now is that the
>> Helper may not have yet received the report before the Leader's request. In
>> this case the Leader would just try again later.
>>
>> Another advantage of this alternative is that it prevents a fairly weak
>> Sybil attack, whereby the Leader makes up `n-1` bogus reports that are
>> processed into the batch.**
>>
>>
>> Thanks for your consideration,
>> Chris P.
>>
>>
>> * Anti-replay is required to prevent privacy violations. See [3].
>> ** The Leader can still mount a slightly stronger attack: Just upload the
>> `n-1` encrypted input shares to the Helper, impersonating the Clients.
>> Though this can be mitigated slightly by rate-limiting requests from the
>> Leader's IP address.
>>
>> [1]
>> https://datatracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#section-6
>> [2] https://github.com/abetterinternet/ppm-specification/issues/130
>> [3]
>> https://datatracker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2
>>
>>

--0000000000000a41bf05d5e24af1
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Hi all, at the end of last week we merged the PR: <a =
href=3D"https://github.com/abetterinternet/ppm-specification/pull/169" targ=
et=3D"_blank">https://github.com/abetterinternet/ppm-specification/pull/169=
</a></div><div></div><div>An issue that we left open is how to prevent aggr=
egators from having to store nonce sets indefinitely. That issue is tracked=
 here: <a href=3D"https://github.com/abetterinternet/ppm-specification/issu=
es/180">https://github.com/abetterinternet/ppm-specification/issues/180</a>=
</div><div><br></div><div>Thanks to all who chime in on this.</div><div><br=
></div><div>Best,</div><div>Chris P.<br></div><br></div><br><div class=3D"g=
mail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Mon, Jan 3, 2022 at 9:=
14 AM Christopher Patton &lt;<a href=3D"mailto:cpatton@cloudflare.com">cpat=
ton@cloudflare.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote=
" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);=
padding-left:1ex"><div dir=3D"ltr">Happy new year friends! The PR with the =
suggested change to anti-replay is up to date and ready for another round o=
f review:<br><div><a href=3D"https://github.com/abetterinternet/ppm-specifi=
cation/pull/169" target=3D"_blank">https://github.com/abetterinternet/ppm-s=
pecification/pull/169</a></div><div><br></div><div> It would be great if, b=
y the end of the week, we can get a sense of whether folks support (some ve=
rsion of) this change or not</div><div><br></div><div>Cheers,</div><div>Chr=
is P.<br> </div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Thu, Dec 2, 2021 at 4:04 PM Christopher Patton &lt;<a hr=
ef=3D"mailto:cpatton@cloudflare.com" target=3D"_blank">cpatton@cloudflare.c=
om</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margi=
n:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex=
"><div dir=3D"ltr"><div>Hi everyone,</div><div><br></div><div>I&#39;ve begu=
n an implementation of the PPM spec with `prio3` [1]. One of the design goa=
ls of the sepc was to make it as easy as possible for a Helper to participa=
te in the protocol, even at the expense of higher resource requirements for=
 the Leader. Based on my experience so far, I think we may have over-indexe=
d on this goal a bit. I propose that we revisit this and consider balancing=
 the work between the Aggregators more evenly. While at it, we may be able =
to resolve [2].</div><div><br></div><div>The spec requires that reports be =
processed in strictly ascending order, where the order is defined in [3]. T=
he ordering is based on a timestamp and random number generated by the clie=
nt and included in the report. I&#39;ll refer to thus (time, rand) pair as =
the nonce. Ordering reports based on the nonce allows the Helper to prevent=
 replay attacks* without needing to store the nonces. In fact, the only lon=
g-term storage it needs for anti-replay is the most recently processed nonc=
e. The downside is that the Leader has to sort reports as they arrive, whic=
h makes horizontally scaling the Leader more challenging.</div><div><br></d=
iv><div>A simpler alternative that we had discussed previously is just havi=
ng each of the Aggregators remember the nonces of the reports they have pro=
cessed. This removes the Leader&#39;s coordination problem, at the cost of =
additional storage requirements. In fact, both Aggregators need to store th=
e set of processed nonces for the duration of a task. This cost can be miti=
gated by refusing to process reports that arrive, say, a day or two late. (=
Of course, this would need to be a parameter that the Aggregators agree on =
in advance.)<br></div><div><br></div><div>Let&#39;s call these schemes opti=
on #1 and option #2 respectively. Previously I was in favor of option #1, b=
ut after getting my hands a little dirty and conferring with some of my col=
leagues, I&#39;ve come to the conclusion that the costs of #1 outweigh the =
costs of #2. I&#39;m curious if others have thought about this and whether =
they are strongly in favor of either option.</div><div><br></div><div>Now, =
if we take option #2, we might also consider revisiting the question raised=
 in [2]. The spec currently requires the Clients to upload their encrypted =
input shares to the Leader, who then piggy-backs the Helper&#39;s input sha=
res on its aggregate requests. Egress fees from cloud providers is typicall=
y higher than ingress, which makes the status quo more expensive to deploy =
than if the Clients uploaded their encrypted input shares directly to each =
Aggregator.</div><div><br></div><div>What might an alternative look like? T=
he client would generate a nonce as usual and upload the nonce and an encry=
pted input share to each of the Aggregators. The aggregation flow would lar=
gely be the same: The Leader sends an Aggregate Request to the Helper, whic=
h contains the nonce of the report it wants to aggregate as usual. The diff=
erence now is that the Helper may not have yet received the report before t=
he Leader&#39;s request. In this case the Leader would just try again later=
.<br></div><div><br></div><div>Another advantage of this alternative is tha=
t it prevents a fairly weak Sybil attack, whereby the Leader makes up `n-1`=
 bogus reports that are processed into the batch.**</div><div><br></div><di=
v><br></div><div>Thanks for your consideration,</div><div>Chris P.<br></div=
><div></div><div><br></div><div><br></div><div>* Anti-replay is required to=
 prevent privacy violations. See [3].</div><div>** The Leader can still mou=
nt a slightly stronger attack: Just upload the=20
`n-1` encrypted input shares to the Helper, impersonating the Clients. Thou=
gh this can be mitigated slightly by rate-limiting requests from the Leader=
&#39;s IP address.<br></div><div><br></div><div>[1] <a href=3D"https://data=
tracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#section-6" target=3D"_b=
lank">https://datatracker.ietf.org/doc/html/draft-patton-cfrg-vdaf-00#secti=
on-6</a></div><div>[2] <a href=3D"https://github.com/abetterinternet/ppm-sp=
ecification/issues/130" target=3D"_blank">https://github.com/abetterinterne=
t/ppm-specification/issues/130</a></div><div>[3] <a href=3D"https://datatra=
cker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2" target=3D"_blank"=
>https://datatracker.ietf.org/doc/html/draft-gpew-priv-ppm#section-4.4.2</a=
></div><div><br></div></div>
</blockquote></div>
</blockquote></div>

--0000000000000a41bf05d5e24af1--


From nobody Tue Jan 18 19:19:55 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id EA5C13A1B53 for <ppm@ietfa.amsl.com>; Tue, 18 Jan 2022 19:19:53 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ZMUKNuip9Yh6 for <ppm@ietfa.amsl.com>; Tue, 18 Jan 2022 19:19:49 -0800 (PST)
Received: from mail-pl1-x629.google.com (mail-pl1-x629.google.com [IPv6:2607:f8b0:4864:20::629]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 8DABE3A1B52 for <ppm@ietf.org>; Tue, 18 Jan 2022 19:19:49 -0800 (PST)
Received: by mail-pl1-x629.google.com with SMTP id u11so840857plh.13 for <ppm@ietf.org>; Tue, 18 Jan 2022 19:19:49 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:from:date:message-id:subject:to; bh=s46Lh25D8hPpj9GBLO5P7pz74jEzq8vDEO0g2f1wGO4=; b=c5403NPWXQbCOS3+cdOoCfXN5DIEyvtChF4aTtfMdS852W1a1/uvKu5NfGLE+NVVp8 7Zz0B17ABJdWAjVe/z1aNQ3pOpoiqLjromezsLlUp4/wSOPwQX+zpjH+7EYCHQUMZB2N P3miaqn8DSRmH/3SrfGtUQRaN4vNRWzYdWOWI=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:from:date:message-id:subject:to; bh=s46Lh25D8hPpj9GBLO5P7pz74jEzq8vDEO0g2f1wGO4=; b=ozXvSsRFspiwAOJ9j0MyYUmglacgAzBTOoz68Ac+Uc6F/jR4k9eIXdjxqvRZ7jetTg jwSlV9X1ChUsL79f6xcAKZ7F/8nFV7CoD0zbGXX3a7lETjBOoxd/T5S6VFqu2CQ5PYP2 ejXSsJn3wmK0NhaU7Cu7bjrUdu50IQUChI0U8Tk4sxXHXD1sTOwykF3IsWKDkIqXxiks iT9tRz/g9bsqpbCtUl9Q5gW6kxQu2ph28gy7PXfJFiPiWjpTknpz2Bkmmqo1QRA17zFt oqLVZK/Kq9yRVCldaTUCREuvpCTRKR9vnePtkfWHcsqio0SMcRABwu9IRFt/Hux9qrj0 nsyA==
X-Gm-Message-State: AOAM5337FyODMeTopCj3BOX+hamSbsWJ1TrOWOaYFjpvh1L2f7ZhkbgL y4K6qcfFVAZfGU/XzvSE0e54tBjhKSc90zFBJgtkt5QkZfXthQ==
X-Google-Smtp-Source: ABdhPJwQh+OqOs7DfDaR8yzKXt5Bto68M1SSfsLvDFLngzX5gvz+xbIjmUB80JXi4/UD3MSQtXvwsLY21pfXMpRkzKA=
X-Received: by 2002:a17:90a:9f81:: with SMTP id o1mr1931903pjp.240.1642562387700;  Tue, 18 Jan 2022 19:19:47 -0800 (PST)
MIME-Version: 1.0
From: Christopher Patton <cpatton@cloudflare.com>
Date: Tue, 18 Jan 2022 19:19:36 -0800
Message-ID: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000a4476905d5e6db3b"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/sCJ1oKR0KMHlZ65FCPUbw3FrGxY>
Subject: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 03:19:54 -0000

--000000000000a4476905d5e6db3b
Content-Type: text/plain; charset="UTF-8"

Hi list,

The current spec has the client upload to the leader a "report" that
contains the encrypted input share of each aggregator. Later, during the
aggregation flow, the leader relays to each helper its share of the report
(i.e., the encrypted input share, the nonce, and extensions). An
alternative upload flow would have the client upload a report share to each
aggregator directly. Both of these approaches have pros and cons, so in
this thread I'd like us to consider which approach makes the most sens or
whether there is a better alternative.

For clarity, the current options are:
Option #1 (status quo): Have the client upload a report to the leader that
contains all of the report shares.
Option #2 (split upload): Have the client upload a report share to each
aggregator directly. For concreteness, this option is spelled out in [1].

There are at least two considerations that lead to the current spec (i.e.,
option #1). One is that option #2 has more failure modes compared to option
#1. For example, if the leader begins aggregating a report before the
helper has received its share from the client, then the report will be
rejected and the leader will be forced to retry later on. The second
consideration is that option #2 has more stringent storage requirements for
the helper, since it has to store its report shares for as long as they can
be aggregated. (Recall that a report may be aggregated multiple times, with
a different aggregation parameter each time [2]. This is necessary to
implement protocols, like Poplar [3], that require the collector to compute
the aggregate result before computing the next aggregation parameter.

The major downside of option #1 (as observed in [4]) is that it increases
the leader-to-helper communication cost. For prio3 the overhead is modest
-- a couple hundred bytes per report share, regardless of the data type --
but for poplar1 the overhead might be significant, since the size of each
input share is linear in the length of the input measurement.

Another downside of option #1 is that, because the leader picks the set of
reports that are aggregated, it can mount a (fairly weak) Sybil attack in
which it stuffs a batch with n-1 bogus reports, thereby leaking the
measurement of a target report entirely. (This requires colluding with the
collector.) Taking option #2 would force the leader to upload the bogus
report shares to the aggregators directly in order to mount the same
attack. In theory, there's not much difference between these two attacks,
since we consider an attacker that controls the network. But as a practical
matter, it would be possible for the helper to mitigate the attack by
rate-limiting requests from the leader's IP address. This isn't perfect,
but it's something. (Note that devising an optional defense against Sybil
attacks is tracked by [5].)

There are, of course, alternatives to these two options. For example:
Option #3 (deployment specific): We could spell the protocol in a way that
leaves this behavior up to the deployment. In particular, for a given task,
the servers would decide whether or not split the uploads.
Option #4 (no re-transmission): The current spec requires the leader to
re-transmit the helper's report share each time the aggregation flow is
executed. Instead, we could have the leader transmit it once and require
the helper to store it thereafter.

Your feedback is greatly appreciated. If anyone has a strong preference for
one of these options, sees a reason why one of them won't work, or has an
idea for an alternative, please chime in :)

Also, it would be great if someone who has implemented IDPFs could tell us
how large the input shares are for, say, inputs of length 64, 128, and 256
bits.

Best,
Chris P.


[1] PR: Draft for split upload:
https://github.com/abetterinternet/ppm-specification/pull/174
[2] Section of VDAF spec defining the aggregation parameter:
https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-preparation
[3] Section of VDAF spec describing Poplar:
https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1
[4] Issue: Cost implications of the leader being the only upload endpoint:
https://github.com/abetterinternet/ppm-specification/issues/130
[5] Issue: Client attestation:
https://github.com/abetterinternet/ppm-specification/issues/89

--000000000000a4476905d5e6db3b
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Hi list,</div><div></div><div></div><div><br></div><d=
iv>The current spec has the client upload to the leader a &quot;report&quot=
; that contains the encrypted input share of each aggregator. Later, during=
 the aggregation flow, the leader relays to each helper its share of the re=
port (i.e., the encrypted input share, the nonce, and extensions). An alter=
native upload flow would have the client upload a report share to each aggr=
egator directly. Both of these approaches have pros and cons, so in this th=
read I&#39;d like us to consider which approach makes the most sens or whet=
her there is a better alternative.<br></div><div><br></div><div>For clarity=
, the current options are:<br></div><div></div><div></div><div></div><div><=
/div><div>Option #1 (status quo): Have the client upload a report to the le=
ader that contains all of the report shares.</div><div>Option #2 (split upl=
oad): Have the client upload a report share to each aggregator directly. Fo=
r concreteness, this option is spelled out in [1].</div><div><br></div><div=
>There are at least two considerations that lead to the current spec (i.e.,=
 option #1). One is that option #2 has more failure modes compared to optio=
n #1. For example, if the leader begins aggregating a report before the hel=
per has received its share from the client, then the report will be rejecte=
d and the leader will be forced to retry later on. The second consideration=
 is that option #2 has more stringent storage requirements for the helper, =
since it has to store its report shares for as long as they can be aggregat=
ed. (Recall that a report may be aggregated multiple times, with a differen=
t=20
aggregation parameter each time [2]. This is necessary to implement=20
protocols, like Poplar [3], that require the collector to compute the=20
aggregate result before computing the next aggregation parameter. </div><di=
v><br></div><div>The major downside of option #1 (as observed in [4]) is th=
at it increases the leader-to-helper communication cost. For prio3 the over=
head is modest -- a couple hundred bytes per report share, regardless of th=
e data type -- but for poplar1 the overhead might be significant, since the=
 size of each input share is linear in the length of the input measurement.=
</div><div><br></div><div>Another downside of option #1 is that, because th=
e leader picks the set of reports that are aggregated, it can mount a (fair=
ly weak) Sybil attack in which it stuffs a batch with n-1 bogus reports, th=
ereby leaking the measurement of a target report entirely. (This requires c=
olluding with the collector.) Taking option #2 would force the leader to up=
load the bogus report shares to the aggregators directly in order to mount =
the same attack. In theory, there&#39;s not much difference between these t=
wo attacks, since we consider an attacker that controls the network. But as=
 a practical matter, it would be possible for the helper to mitigate the at=
tack by rate-limiting requests from the leader&#39;s IP address. This isn&#=
39;t perfect, but it&#39;s something. (Note that devising an optional defen=
se against Sybil attacks is tracked by [5].)<br></div><div><br></div><div>T=
here are, of course, alternatives to these two options. For example:<br><di=
v>Option #3 (deployment specific): We could spell the protocol in a=20
way that leaves this behavior up to the deployment. In particular, for a
 given task, the servers would decide whether or not split the uploads.<br>=
</div><div>Option #4 (no re-transmission):=20
The current spec requires the leader to re-transmit the helper&#39;s report=
=20
share each time the aggregation flow is executed. Instead, we could have
 the leader transmit it once and require the helper to store it=20
thereafter.</div><div><br></div><div>Your feedback is greatly appreciated. =
If anyone has a strong preference for one of these options, sees a reason w=
hy one of them won&#39;t work, or has an idea for an alternative, please ch=
ime in :)</div><div><br></div><div> Also, it would be great if someone who =
has implemented IDPFs could tell us how large the input shares are for, say=
, inputs of length 64, 128, and 256 bits.</div><div><br></div><div>Best, <b=
r></div><div>Chris P.<br></div></div><br><div><br></div><div>[1] PR: Draft =
for split upload: <a href=3D"https://github.com/abetterinternet/ppm-specifi=
cation/pull/174" target=3D"_blank">https://github.com/abetterinternet/ppm-s=
pecification/pull/174</a></div><div></div><div>[2] Section of VDAF spec def=
ining the aggregation parameter: <a href=3D"https://cjpatton.github.io/vdaf=
/draft-patton-cfrg-vdaf.html#name-preparation">https://cjpatton.github.io/v=
daf/draft-patton-cfrg-vdaf.html#name-preparation</a></div><div>[3] Section =
of VDAF spec describing Poplar: <a href=3D"https://cjpatton.github.io/vdaf/=
draft-patton-cfrg-vdaf.html#name-poplar1">https://cjpatton.github.io/vdaf/d=
raft-patton-cfrg-vdaf.html#name-poplar1</a></div><div></div><div>[4] Issue:=
 Cost implications of the leader being the only upload endpoint: <a href=3D=
"https://github.com/abetterinternet/ppm-specification/issues/130" target=3D=
"_blank">https://github.com/abetterinternet/ppm-specification/issues/130</a=
></div><div>[5] Issue: Client attestation: <a href=3D"https://github.com/ab=
etterinternet/ppm-specification/issues/89">https://github.com/abetterintern=
et/ppm-specification/issues/89</a></div><div><br></div></div>

--000000000000a4476905d5e6db3b--


From nobody Wed Jan 19 07:15:26 2022
Return-Path: <csharrison@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 2CF1D3A0E56 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 07:15:23 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id H05ez_thTzsi for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 07:15:21 -0800 (PST)
Received: from mail-ed1-x535.google.com (mail-ed1-x535.google.com [IPv6:2a00:1450:4864:20::535]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id DBB013A0E60 for <ppm@ietf.org>; Wed, 19 Jan 2022 07:13:27 -0800 (PST)
Received: by mail-ed1-x535.google.com with SMTP id f17so5076594edw.2 for <ppm@ietf.org>; Wed, 19 Jan 2022 07:13:27 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=wr1d1yAifPtFGhuxku/hLXZE6bt35a9mYscoVw6HuC4=; b=LnaksFZeaEJ3fC29MIUSheNOK1I2fU0L9uwAMjs5onUvtoOa6CFDwr6P7kiVwVYoPA b7Q3x0PJSnVQDGbh+7+0FAj3IMoJ8wFoVjKX2Hm+8BlS1U/DNLmwC3479UySjAV0Oa9N S4QtzeYHYPa4bRp2C6NZGvyjvKP0TAa1THaHs8Btj7NcbFlRR/D/eMk5IEPNxCAIr2NE oxBGcMzS2qXmB5thPye5Qx9v/swGndn8onWRu88Fo92wV+6krxcHDigisT9+FMzhT/Mm Z5ldGHwgCiIeAPB9T71GZL89PuzQrBCmThAEQSVZhC/b2Qq3l8788ec+WNbMi3kbk1Yi +Cgw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=wr1d1yAifPtFGhuxku/hLXZE6bt35a9mYscoVw6HuC4=; b=CTvAT415m04kxTOQ7XYKRpjbdqNnfCaprB5b15yP4jetGXDW674E5sNgb9MTGgqix4 sv34UouFIqmesd1NcRZS1hEy2q1ePoqEs5GV/5ymGfvQEhfCETREOk56unFJdXtw7gnw DCJ/lCcQXiCNOmkVxqAC8bjsx8f/2zS5ZO4CKiwxXx4el8zvU/YBu0zkrFvjiFyYzb9v JusR1rnSMfZvbTJqyY7STQgfN04XusAPxcMzEuY7TxXjz4FlbbCZD6ooIsUPhXvD8LJA Vb8jAGknJFMi49HpsnBGOcBwoSR+IzndKfvPQLyNUFA+u9xT1gQCFKl51qW37SkROayn 2Mbw==
X-Gm-Message-State: AOAM530shdLMI8yf/Q0UuRR2ATubLB01PQwGhYEGknr3trGX2+lRunfg XJ+l4FT2EhTmCZMFT5CWcXKvvdK6lN0sVZWoTmR75w==
X-Google-Smtp-Source: ABdhPJzNVDgyByXLL8058Y+O5ItvihuhU55l0F3GKWtjfqup91Z/75Oea1VsvoG25ev3RbuR3JLZ3ZSAkSnFgYQZ1Z4=
X-Received: by 2002:a05:6402:1258:: with SMTP id l24mr25533969edw.163.1642605204786;  Wed, 19 Jan 2022 07:13:24 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com>
In-Reply-To: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com>
From: Charlie Harrison <csharrison@google.com>
Date: Wed, 19 Jan 2022 09:13:13 -0600
Message-ID: <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com>
To: Christopher Patton <cpatton=40cloudflare.com@dmarc.ietf.org>
Cc: ppm@ietf.org, Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>
Content-Type: multipart/alternative; boundary="000000000000bd7fd105d5f0d366"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/j9TLBPLpQCMxnu9aLvgqy4VWBWk>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 15:15:23 -0000

--000000000000bd7fd105d5f0d366
Content-Type: text/plain; charset="UTF-8"

Thanks for prompting this discussion, Christopher. I would appreciate your
thoughts on Option 3, since it seems like there are very good reasons to
prefer either Option 1 or 2.

Another idle concern I had, which I cannot substantiate immediately, is
whether Option #1 meaningfully limits the space of possible MPC protocols
we could support, by revealing the # of reports being sent to the helper to
the leader.

cc schoppman who I think can give numbers for input shares in the current
C++ implementation, as well as rkubis who likely has opinions here.

On Tue, Jan 18, 2022 at 9:20 PM Christopher Patton <cpatton=
40cloudflare.com@dmarc.ietf.org> wrote:

> Hi list,
>
> The current spec has the client upload to the leader a "report" that
> contains the encrypted input share of each aggregator. Later, during the
> aggregation flow, the leader relays to each helper its share of the report
> (i.e., the encrypted input share, the nonce, and extensions). An
> alternative upload flow would have the client upload a report share to each
> aggregator directly. Both of these approaches have pros and cons, so in
> this thread I'd like us to consider which approach makes the most sens or
> whether there is a better alternative.
>
> For clarity, the current options are:
> Option #1 (status quo): Have the client upload a report to the leader that
> contains all of the report shares.
> Option #2 (split upload): Have the client upload a report share to each
> aggregator directly. For concreteness, this option is spelled out in [1].
>
> There are at least two considerations that lead to the current spec (i.e.,
> option #1). One is that option #2 has more failure modes compared to option
> #1. For example, if the leader begins aggregating a report before the
> helper has received its share from the client, then the report will be
> rejected and the leader will be forced to retry later on. The second
> consideration is that option #2 has more stringent storage requirements for
> the helper, since it has to store its report shares for as long as they can
> be aggregated. (Recall that a report may be aggregated multiple times, with
> a different aggregation parameter each time [2]. This is necessary to
> implement protocols, like Poplar [3], that require the collector to compute
> the aggregate result before computing the next aggregation parameter.
>
> The major downside of option #1 (as observed in [4]) is that it increases
> the leader-to-helper communication cost. For prio3 the overhead is modest
> -- a couple hundred bytes per report share, regardless of the data type --
> but for poplar1 the overhead might be significant, since the size of each
> input share is linear in the length of the input measurement.
>

I don't quite understand this. I thought in Prio the input shares were
linear in the length of the input measurement even excluding the proofs,
since they are simple vector shares. Are you talking about just the SNIP
proofs?


>
> Another downside of option #1 is that, because the leader picks the set of
> reports that are aggregated, it can mount a (fairly weak) Sybil attack in
> which it stuffs a batch with n-1 bogus reports, thereby leaking the
> measurement of a target report entirely. (This requires colluding with the
> collector.) Taking option #2 would force the leader to upload the bogus
> report shares to the aggregators directly in order to mount the same
> attack. In theory, there's not much difference between these two attacks,
> since we consider an attacker that controls the network. But as a practical
> matter, it would be possible for the helper to mitigate the attack by
> rate-limiting requests from the leader's IP address. This isn't perfect,
> but it's something. (Note that devising an optional defense against Sybil
> attacks is tracked by [5].)
>
> There are, of course, alternatives to these two options. For example:
> Option #3 (deployment specific): We could spell the protocol in a way that
> leaves this behavior up to the deployment. In particular, for a given task,
> the servers would decide whether or not split the uploads.
> Option #4 (no re-transmission): The current spec requires the leader to
> re-transmit the helper's report share each time the aggregation flow is
> executed. Instead, we could have the leader transmit it once and require
> the helper to store it thereafter.
>
> Your feedback is greatly appreciated. If anyone has a strong preference
> for one of these options, sees a reason why one of them won't work, or has
> an idea for an alternative, please chime in :)
>
> Also, it would be great if someone who has implemented IDPFs could tell us
> how large the input shares are for, say, inputs of length 64, 128, and 256
> bits.
>
> Best,
> Chris P.
>
>
> [1] PR: Draft for split upload:
> https://github.com/abetterinternet/ppm-specification/pull/174
> [2] Section of VDAF spec defining the aggregation parameter:
> https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-preparation
> [3] Section of VDAF spec describing Poplar:
> https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1
> [4] Issue: Cost implications of the leader being the only upload endpoint:
> https://github.com/abetterinternet/ppm-specification/issues/130
> [5] Issue: Client attestation:
> https://github.com/abetterinternet/ppm-specification/issues/89
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000bd7fd105d5f0d366
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><div>Thanks for prompting this discussion=
, Christopher. I would appreciate your thoughts on Option 3, since it seems=
 like there are very good reasons to prefer either Option 1 or 2.</div><div=
><br></div><div>Another idle concern I had, which I cannot substantiate imm=
ediately, is whether Option #1 meaningfully limits the space of possible MP=
C protocols we could support, by revealing the # of reports being sent to t=
he helper to the leader.</div><div><br></div><div>cc schoppman who I think =
can give numbers for input shares in the current C++ implementation, as wel=
l as rkubis who likely has opinions here.</div></div><br><div class=3D"gmai=
l_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, Jan 18, 2022 at 9:20=
 PM Christopher Patton &lt;cpatton=3D<a href=3D"mailto:40cloudflare.com@dma=
rc.ietf.org">40cloudflare.com@dmarc.ietf.org</a>&gt; wrote:<br></div><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1=
px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>Hi list,<=
/div><div></div><div></div><div><br></div><div>The current spec has the cli=
ent upload to the leader a &quot;report&quot; that contains the encrypted i=
nput share of each aggregator. Later, during the aggregation flow, the lead=
er relays to each helper its share of the report (i.e., the encrypted input=
 share, the nonce, and extensions). An alternative upload flow would have t=
he client upload a report share to each aggregator directly. Both of these =
approaches have pros and cons, so in this thread I&#39;d like us to conside=
r which approach makes the most sens or whether there is a better alternati=
ve.<br></div><div><br></div><div>For clarity, the current options are:<br><=
/div><div></div><div></div><div></div><div></div><div>Option #1 (status quo=
): Have the client upload a report to the leader that contains all of the r=
eport shares.</div><div>Option #2 (split upload): Have the client upload a =
report share to each aggregator directly. For concreteness, this option is =
spelled out in [1].</div><div><br></div><div>There are at least two conside=
rations that lead to the current spec (i.e., option #1). One is that option=
 #2 has more failure modes compared to option #1. For example, if the leade=
r begins aggregating a report before the helper has received its share from=
 the client, then the report will be rejected and the leader will be forced=
 to retry later on. The second consideration is that option #2 has more str=
ingent storage requirements for the helper, since it has to store its repor=
t shares for as long as they can be aggregated. (Recall that a report may b=
e aggregated multiple times, with a different=20
aggregation parameter each time [2]. This is necessary to implement=20
protocols, like Poplar [3], that require the collector to compute the=20
aggregate result before computing the next aggregation parameter. </div><di=
v><br></div><div>The major downside of option #1 (as observed in [4]) is th=
at it increases the leader-to-helper communication cost. For prio3 the over=
head is modest -- a couple hundred bytes per report share, regardless of th=
e data type -- but for poplar1 the overhead might be significant, since the=
 size of each input share is linear in the length of the input measurement.=
</div></div></blockquote><div><br></div><div>I don&#39;t quite understand t=
his. I thought in Prio the input shares were linear in the length of the in=
put measurement even excluding the proofs, since they are simple vector sha=
res. Are you talking about just the SNIP proofs?=C2=A0</div><div>=C2=A0</di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><=
br></div><div>Another downside of option #1 is that, because the leader pic=
ks the set of reports that are aggregated, it can mount a (fairly weak) Syb=
il attack in which it stuffs a batch with n-1 bogus reports, thereby leakin=
g the measurement of a target report entirely. (This requires colluding wit=
h the collector.) Taking option #2 would force the leader to upload the bog=
us report shares to the aggregators directly in order to mount the same att=
ack. In theory, there&#39;s not much difference between these two attacks, =
since we consider an attacker that controls the network. But as a practical=
 matter, it would be possible for the helper to mitigate the attack by rate=
-limiting requests from the leader&#39;s IP address. This isn&#39;t perfect=
, but it&#39;s something. (Note that devising an optional defense against S=
ybil attacks is tracked by [5].)<br></div><div><br></div><div>There are, of=
 course, alternatives to these two options. For example:<br><div>Option #3 =
(deployment specific): We could spell the protocol in a=20
way that leaves this behavior up to the deployment. In particular, for a
 given task, the servers would decide whether or not split the uploads.<br>=
</div><div>Option #4 (no re-transmission):=20
The current spec requires the leader to re-transmit the helper&#39;s report=
=20
share each time the aggregation flow is executed. Instead, we could have
 the leader transmit it once and require the helper to store it=20
thereafter.</div><div><br></div><div>Your feedback is greatly appreciated. =
If anyone has a strong preference for one of these options, sees a reason w=
hy one of them won&#39;t work, or has an idea for an alternative, please ch=
ime in :)</div><div><br></div><div> Also, it would be great if someone who =
has implemented IDPFs could tell us how large the input shares are for, say=
, inputs of length 64, 128, and 256 bits.</div><div><br></div><div>Best, <b=
r></div><div>Chris P.<br></div></div><br><div><br></div><div>[1] PR: Draft =
for split upload: <a href=3D"https://github.com/abetterinternet/ppm-specifi=
cation/pull/174" target=3D"_blank">https://github.com/abetterinternet/ppm-s=
pecification/pull/174</a></div><div></div><div>[2] Section of VDAF spec def=
ining the aggregation parameter: <a href=3D"https://cjpatton.github.io/vdaf=
/draft-patton-cfrg-vdaf.html#name-preparation" target=3D"_blank">https://cj=
patton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-preparation</a></div=
><div>[3] Section of VDAF spec describing Poplar: <a href=3D"https://cjpatt=
on.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1" target=3D"_blan=
k">https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1=
</a></div><div></div><div>[4] Issue: Cost implications of the leader being =
the only upload endpoint: <a href=3D"https://github.com/abetterinternet/ppm=
-specification/issues/130" target=3D"_blank">https://github.com/abetterinte=
rnet/ppm-specification/issues/130</a></div><div>[5] Issue: Client attestati=
on: <a href=3D"https://github.com/abetterinternet/ppm-specification/issues/=
89" target=3D"_blank">https://github.com/abetterinternet/ppm-specification/=
issues/89</a></div><div><br></div></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div>

--000000000000bd7fd105d5f0d366--


From nobody Wed Jan 19 09:00:24 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 305B13A13CA for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 09:00:23 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.087
X-Spam-Level: 
X-Spam-Status: No, score=-7.087 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_KAM_HTML_FONT_INVALID=0.01, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id eanS54XK-P0O for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 09:00:19 -0800 (PST)
Received: from mail-ed1-x535.google.com (mail-ed1-x535.google.com [IPv6:2a00:1450:4864:20::535]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 7616D3A0899 for <ppm@ietf.org>; Wed, 19 Jan 2022 09:00:18 -0800 (PST)
Received: by mail-ed1-x535.google.com with SMTP id cx27so15413918edb.1 for <ppm@ietf.org>; Wed, 19 Jan 2022 09:00:18 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=Hbf3bVsagzNjiDF5UGyB5r8nG6oNlcrWwJeFn1M2DZ0=; b=OVhPH/XKXKBfhARnNBcmwGyotkf42ru/iXJpINlBxhKcycBzgjjk6PWOXktjdvUrNk YXI3oqmmUa1OjaBDraUxUr15RqF7mEsyZKZEDQTokUzYDPGEURlLiEE5AB3EIMCewACO 9G1r5hcnElVW7vitRCn/SieBDg1ZttVRXpj967XkCFQtqnzl6hwV41jM8NH2IW4vENYJ e4Ftz/F+51husaW1+HzYTeX18PHt9yAsB/DZWMkvdpUjaLJDL79Mi42JaloQl30VwobT Hr5yNEPO3vvuBvxi2v/IC3RbXrExglcE0DGlkzAhObhirKSGJRvvIJFMyy0CmXE6ac+M tzlA==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=Hbf3bVsagzNjiDF5UGyB5r8nG6oNlcrWwJeFn1M2DZ0=; b=1WkQL+Rg4Nn9CgSjh4lpSaNOGbVceDzcS34NkK0+z/0kIBV4hjUsLxh90M+W1dS63F KeSpkaV7PDWqrC/uPL4D2QyPVRCuX3ROvTPpgdx8K5dXinoPGHsHx/aNfKMmtGgenYhu 6hrDGLyX9r1k7hO4qqOYLslJ7MDqIfwXMv2VmGHh5rWMfB5B3FVv8cUtt/UE4wm8GWji WdwgB4jQ9fGbA5OrwvhvK0gkIo9CIzj1kKfbNT8h9d1lVPN47SICwMzlLGXCxfSmDi// iD17bLuSkJ7vrLm+4YfvbJSXGLHfvLzp8JhBKvhQXKw3Y5jU6iideah2HoJ8BeW4Ac6F ESyA==
X-Gm-Message-State: AOAM530runZSaokQV2FR6Pi4LhOiGNPyA+gQS7mki+6/TlHA+g+e01UI u88qHlLhlMNaeyoRXTI8oPCYWiDTRsfmS4nCzeE=
X-Google-Smtp-Source: ABdhPJwx5kFqSg13ViUuA68eVYhsu2iUrMHHdqCUQecoS+8pxDcKFAHhhbif+Uwc9A0m3isqlOeLxsm1garszbhRjW8=
X-Received: by 2002:a05:6512:ea7:: with SMTP id bi39mr27668964lfb.312.1642611613401;  Wed, 19 Jan 2022 09:00:13 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com>
In-Reply-To: <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Wed, 19 Jan 2022 09:00:02 -0800
Message-ID: <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com>
To: Charlie Harrison <csharrison=40google.com@dmarc.ietf.org>
Cc: Christopher Patton <cpatton=40cloudflare.com@dmarc.ietf.org>,  Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>,  ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000b8890905d5f251e1"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/DbZSHHxapY37qanovDB_bN0a5lo>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 17:00:23 -0000

--000000000000b8890905d5f251e1
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Thanks for kicking this off Chris, and I appreciate the detailed analysis.

Ultimately, I would argue this should be a choice made in the
implementation, and not in the specification. A bunch more thoughts below:

There is a 4th option being discussed in issue #166
<https://github.com/abetterinternet/ppm-specification/issues/166> [1],
which moves in the opposite direction by proposing sending reports to a new
party called a consumer which is entirely untrusted and specified by the
collector. Note that the consumer would also also be able to mount the
Sybil attack, which I=E2=80=99ll address below.

This is the architecture proposed in Interoperable Private Attribution
<https://docs.google.com/document/d/1KpdSKD8-Rn0bWPTu4UtK54ks0yv2j22pA5SrAD=
9av4s/edit#heading=3Dh.f4x9f0nqv28x>
[2], which we intend to be an implementation of PPM. I believe that it is
also the Multi-Browser Aggregation Service Explainer
<https://github.com/WICG/conversion-measurement-api/blob/main/SERVICE.md>
[3], with reports being sent to the reporting origin, but I would look to
Charlie Harrison to confirm.

The rationale here is almost entirely operational. Both of the APIs above
are implemented within an user-agent, but are used by arbitrary websites
and apps. Consuming and storing these reports sent from instances of those
user-agents is a non-trivial infrastructure challenge, and we should assume
that aggregators will not have 100% uptime. In fact, improving from 99.9%
uptime to 99.99% uptime can often mean an order of magnitude increase in
operational costs.

I would assume that the set of parties offering their services as
aggregators will be fairly constrained for the near-to-mid term, and as
such, collectors will have minimal choice when it comes to finding the
appropriate trade offs between operational concerns like uptime vs
infrastructure costs. As such, I believe it=E2=80=99s beneficial to move as=
 much of
this cost and operational effort outside of the aggregators.

To the point on Sybil attacks: nearly all the APIs that would implement PPM
will run in user-agents on untrusted and unknown devices. The code used to
generate those reports will be open source, or at least the protocol will
be publicly known, so the bigger threat, in my opinion, is a malicious
entity operating a modified user-agent to generate events which enable a
Sybil attack. The IPA proposal linked above (as well as the Attribution
Reporting API with Aggregate Reports
<https://github.com/WICG/conversion-measurement-api/blob/main/AGGREGATE.md>
[4]) propose using differential privacy to address these types of attacks.
This approach should also mitigate Sybil attacks.

I don=E2=80=99t believe that all implementations will end up using differen=
tial
privacy, and thus may find the proposal for reports to be sent directly to
all aggregators useful, but for those that do implement this mitigation,
sending the reports directly to all aggregators presents some meaningful
challenges.

Finally, in terms of cost, given that the majority of the bandwidth cost is
in egress cost (i.e. leaving a cloud provider network.) I would agree that
it would be useful to be able to specify multiple locations for reports to
be sent, somewhat similar to the structure proposed in this comment
<https://github.com/abetterinternet/ppm-specification/issues/130#issuecomme=
nt-967730879>
[5]. This would enable reports to be sent directly from clients into
different cloud providers=E2=80=93and the server running in those cloud pro=
viders
could be a proxy of the consumer (proposed above), a proxy of the leader,
or the aggregators themselves. This structure would allow us to separate
the egress cost concerns from the Sybil attack concerns.


Cheers,

Erik

[1] Issue 166: adding an untrusted consumer that stores reports before
aggregation: https://github.com/abetterinternet/ppm-specification/issues/16=
6

[2] Interoperable Private Attribution proposal:
https://docs.google.com/document/d/1KpdSKD8-Rn0bWPTu4UtK54ks0yv2j22pA5SrAD9=
av4s/edit#heading=3Dh.f4x9f0nqv28x

[3] Multi-Browser Aggregation Service Explainer, subset of the Attribution
Reporting API proposal:
https://github.com/WICG/conversion-measurement-api/blob/main/SERVICE.md

[4] Attribution Reporting API with Aggregate Reports, subset of the
Attribution Reporting API proposal:
https://github.com/WICG/conversion-measurement-api/blob/main/AGGREGATE.md

[5] Issue 130 comment, proposing proxies to receive reports in separate
cloud providers:
https://github.com/abetterinternet/ppm-specification/issues/130#issuecommen=
t-967730879


On Wed, Jan 19, 2022 at 7:15 AM Charlie Harrison <csharrison=3D
40google.com@dmarc.ietf.org> wrote:

> Thanks for prompting this discussion, Christopher. I would appreciate you=
r
> thoughts on Option 3, since it seems like there are very good reasons to
> prefer either Option 1 or 2.
>
> Another idle concern I had, which I cannot substantiate immediately, is
> whether Option #1 meaningfully limits the space of possible MPC protocols
> we could support, by revealing the # of reports being sent to the helper =
to
> the leader.
>
> cc schoppman who I think can give numbers for input shares in the current
> C++ implementation, as well as rkubis who likely has opinions here.
>
> On Tue, Jan 18, 2022 at 9:20 PM Christopher Patton <cpatton=3D
> 40cloudflare.com@dmarc.ietf.org> wrote:
>
>> Hi list,
>>
>> The current spec has the client upload to the leader a "report" that
>> contains the encrypted input share of each aggregator. Later, during the
>> aggregation flow, the leader relays to each helper its share of the repo=
rt
>> (i.e., the encrypted input share, the nonce, and extensions). An
>> alternative upload flow would have the client upload a report share to e=
ach
>> aggregator directly. Both of these approaches have pros and cons, so in
>> this thread I'd like us to consider which approach makes the most sens o=
r
>> whether there is a better alternative.
>>
>> For clarity, the current options are:
>> Option #1 (status quo): Have the client upload a report to the leader
>> that contains all of the report shares.
>> Option #2 (split upload): Have the client upload a report share to each
>> aggregator directly. For concreteness, this option is spelled out in [1]=
.
>>
>> There are at least two considerations that lead to the current spec
>> (i.e., option #1). One is that option #2 has more failure modes compared=
 to
>> option #1. For example, if the leader begins aggregating a report before
>> the helper has received its share from the client, then the report will =
be
>> rejected and the leader will be forced to retry later on. The second
>> consideration is that option #2 has more stringent storage requirements =
for
>> the helper, since it has to store its report shares for as long as they =
can
>> be aggregated. (Recall that a report may be aggregated multiple times, w=
ith
>> a different aggregation parameter each time [2]. This is necessary to
>> implement protocols, like Poplar [3], that require the collector to comp=
ute
>> the aggregate result before computing the next aggregation parameter.
>>
>> The major downside of option #1 (as observed in [4]) is that it increase=
s
>> the leader-to-helper communication cost. For prio3 the overhead is modes=
t
>> -- a couple hundred bytes per report share, regardless of the data type =
--
>> but for poplar1 the overhead might be significant, since the size of eac=
h
>> input share is linear in the length of the input measurement.
>>
>
> I don't quite understand this. I thought in Prio the input shares were
> linear in the length of the input measurement even excluding the proofs,
> since they are simple vector shares. Are you talking about just the SNIP
> proofs?
>
>
>>
>> Another downside of option #1 is that, because the leader picks the set
>> of reports that are aggregated, it can mount a (fairly weak) Sybil attac=
k
>> in which it stuffs a batch with n-1 bogus reports, thereby leaking the
>> measurement of a target report entirely. (This requires colluding with t=
he
>> collector.) Taking option #2 would force the leader to upload the bogus
>> report shares to the aggregators directly in order to mount the same
>> attack. In theory, there's not much difference between these two attacks=
,
>> since we consider an attacker that controls the network. But as a practi=
cal
>> matter, it would be possible for the helper to mitigate the attack by
>> rate-limiting requests from the leader's IP address. This isn't perfect,
>> but it's something. (Note that devising an optional defense against Sybi=
l
>> attacks is tracked by [5].)
>>
>> There are, of course, alternatives to these two options. For example:
>> Option #3 (deployment specific): We could spell the protocol in a way
>> that leaves this behavior up to the deployment. In particular, for a giv=
en
>> task, the servers would decide whether or not split the uploads.
>> Option #4 (no re-transmission): The current spec requires the leader to
>> re-transmit the helper's report share each time the aggregation flow is
>> executed. Instead, we could have the leader transmit it once and require
>> the helper to store it thereafter.
>>
>> Your feedback is greatly appreciated. If anyone has a strong preference
>> for one of these options, sees a reason why one of them won't work, or h=
as
>> an idea for an alternative, please chime in :)
>>
>> Also, it would be great if someone who has implemented IDPFs could tell
>> us how large the input shares are for, say, inputs of length 64, 128, an=
d
>> 256 bits.
>>
>> Best,
>> Chris P.
>>
>>
>> [1] PR: Draft for split upload:
>> https://github.com/abetterinternet/ppm-specification/pull/174
>> [2] Section of VDAF spec defining the aggregation parameter:
>> https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-prepara=
tion
>> [3] Section of VDAF spec describing Poplar:
>> https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1
>> [4] Issue: Cost implications of the leader being the only upload
>> endpoint: https://github.com/abetterinternet/ppm-specification/issues/13=
0
>> [5] Issue: Client attestation:
>> https://github.com/abetterinternet/ppm-specification/issues/89
>>
>> --
>> Ppm mailing list
>> Ppm@ietf.org
>> https://www.ietf.org/mailman/listinfo/ppm
>>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000b8890905d5f251e1
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><span id=3D"gmail-docs-internal-guid-7091e80b-7fff-7c0e-b2=
b6-16a440521498"><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;ma=
rgin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(=
0,0,0);background-color:transparent;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Thanks fo=
r kicking this off Chris, and I appreciate the detailed analysis.</span></p=
><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:=
0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);backg=
round-color:transparent;font-variant-numeric:normal;font-variant-east-asian=
:normal;vertical-align:baseline;white-space:pre-wrap">Ultimately, I would a=
rgue this should be a choice made in the implementation, and not in the spe=
cification. A bunch more thoughts below:</span></p><br><p dir=3D"ltr" style=
=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-=
size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;f=
ont-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:ba=
seline;white-space:pre-wrap">There is a 4th option being discussed in </spa=
n><a href=3D"https://github.com/abetterinternet/ppm-specification/issues/16=
6" style=3D"text-decoration-line:none"><span style=3D"font-size:10pt;font-f=
amily:Arial;background-color:transparent;font-variant-numeric:normal;font-v=
ariant-east-asian:normal;text-decoration-line:underline;vertical-align:base=
line;white-space:pre-wrap">issue #166</span></a><span style=3D"font-size:10=
pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-var=
iant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;=
white-space:pre-wrap"> [1], which moves in the opposite direction by propos=
ing sending reports to a new party called a </span><span style=3D"font-size=
:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-=
style:italic;font-variant-numeric:normal;font-variant-east-asian:normal;ver=
tical-align:baseline;white-space:pre-wrap">consumer</span><span style=3D"fo=
nt-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparen=
t;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align=
:baseline;white-space:pre-wrap"> which is entirely untrusted and specified =
by the </span><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0=
,0);background-color:transparent;font-style:italic;font-variant-numeric:nor=
mal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-=
wrap">collector</span><span style=3D"font-size:10pt;font-family:Arial;color=
:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-v=
ariant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">. No=
te that the </span><span style=3D"font-size:10pt;font-family:Arial;color:rg=
b(0,0,0);background-color:transparent;font-style:italic;font-variant-numeri=
c:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space=
:pre-wrap">consumer</span><span style=3D"font-size:10pt;font-family:Arial;c=
olor:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;fo=
nt-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">=
 would also also be able to mount the Sybil attack, which I=E2=80=99ll addr=
ess below.</span></p><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-to=
p:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;co=
lor:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;fon=
t-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">T=
his is the architecture proposed in </span><a href=3D"https://docs.google.c=
om/document/d/1KpdSKD8-Rn0bWPTu4UtK54ks0yv2j22pA5SrAD9av4s/edit#heading=3Dh=
.f4x9f0nqv28x" style=3D"text-decoration-line:none"><span style=3D"font-size=
:10pt;font-family:Arial;background-color:transparent;font-style:italic;font=
-variant-numeric:normal;font-variant-east-asian:normal;text-decoration-line=
:underline;vertical-align:baseline;white-space:pre-wrap">Interoperable Priv=
ate Attribution</span></a><span style=3D"font-size:10pt;font-family:Arial;c=
olor:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;fo=
nt-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">=
 [2], which we intend to be an implementation of PPM. I believe that it is =
also the </span><a href=3D"https://github.com/WICG/conversion-measurement-a=
pi/blob/main/SERVICE.md" style=3D"text-decoration-line:none"><span style=3D=
"font-size:10pt;font-family:Arial;background-color:transparent;font-style:i=
talic;font-variant-numeric:normal;font-variant-east-asian:normal;text-decor=
ation-line:underline;vertical-align:baseline;white-space:pre-wrap">Multi-Br=
owser Aggregation Service Explainer</span></a><span style=3D"font-size:10pt=
;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-varia=
nt-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;wh=
ite-space:pre-wrap"> [3], with reports being sent to the </span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-style:italic;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">reporting origin</s=
pan><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);backgr=
ound-color:transparent;font-variant-numeric:normal;font-variant-east-asian:=
normal;vertical-align:baseline;white-space:pre-wrap">, but I would look to =
Charlie Harrison to confirm.</span></p><br><p dir=3D"ltr" style=3D"line-hei=
ght:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;fo=
nt-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-=
numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white=
-space:pre-wrap">The rationale here is almost entirely operational. Both of=
 the APIs above are implemented within an user-agent, but are used by arbit=
rary websites and apps. Consuming and storing these reports sent from insta=
nces of those user-agents is a non-trivial infrastructure challenge, and we=
 should assume that </span><span style=3D"font-size:10pt;font-family:Arial;=
color:rgb(0,0,0);background-color:transparent;font-style:italic;font-varian=
t-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;whi=
te-space:pre-wrap">aggregators</span><span style=3D"font-size:10pt;font-fam=
ily:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeri=
c:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space=
:pre-wrap"> will not have 100% uptime. In fact, improving from 99.9% uptime=
 to 99.99% uptime can often mean an order of magnitude increase in operatio=
nal costs.</span></p><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-to=
p:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;co=
lor:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;fon=
t-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">I=
 would assume that the set of parties offering their services as </span><sp=
an style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-co=
lor:transparent;font-style:italic;font-variant-numeric:normal;font-variant-=
east-asian:normal;vertical-align:baseline;white-space:pre-wrap">aggregators=
</span><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);bac=
kground-color:transparent;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap"> will be fairly con=
strained for the near-to-mid term, and as such, </span><span style=3D"font-=
size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;f=
ont-style:italic;font-variant-numeric:normal;font-variant-east-asian:normal=
;vertical-align:baseline;white-space:pre-wrap">collectors</span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical=
-align:baseline;white-space:pre-wrap"> will have minimal choice when it com=
es to finding the appropriate trade offs between operational concerns like =
uptime vs infrastructure costs. As such, I believe it=E2=80=99s beneficial =
to move as much of this cost and operational effort outside of the </span><=
span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-=
color:transparent;font-style:italic;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">aggregato=
rs</span><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);b=
ackground-color:transparent;font-variant-numeric:normal;font-variant-east-a=
sian:normal;vertical-align:baseline;white-space:pre-wrap">.</span></p><br><=
p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><=
span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-=
color:transparent;font-variant-numeric:normal;font-variant-east-asian:norma=
l;vertical-align:baseline;white-space:pre-wrap">To the point on Sybil attac=
ks: nearly all the APIs that would implement PPM will run in user-agents on=
 untrusted and unknown devices. The code used to generate those reports wil=
l be open source, or at least the protocol will be publicly known, so the b=
igger threat, in my opinion, is a malicious entity operating a modified use=
r-agent to generate events which enable a Sybil attack. The IPA proposal li=
nked above (as well as the </span><a href=3D"https://github.com/WICG/conver=
sion-measurement-api/blob/main/AGGREGATE.md" style=3D"text-decoration-line:=
none"><span style=3D"font-size:10pt;font-family:Arial;background-color:tran=
sparent;font-variant-numeric:normal;font-variant-east-asian:normal;text-dec=
oration-line:underline;vertical-align:baseline;white-space:pre-wrap">Attrib=
ution Reporting API with Aggregate Reports</span></a><span style=3D"font-si=
ze:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;fon=
t-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:base=
line;white-space:pre-wrap"> [4]) propose using differential privacy to addr=
ess these types of attacks. This approach should also mitigate Sybil attack=
s.</span></p><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;ma=
rgin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(=
0,0,0);background-color:transparent;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">I don=E2=
=80=99t believe that all implementations will end up using differential pri=
vacy, and thus may find the proposal for reports to be sent directly to all=
 </span><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);ba=
ckground-color:transparent;font-style:italic;font-variant-numeric:normal;fo=
nt-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">=
aggregators</span><span style=3D"font-size:10pt;font-family:Arial;color:rgb=
(0,0,0);background-color:transparent;font-variant-numeric:normal;font-varia=
nt-east-asian:normal;vertical-align:baseline;white-space:pre-wrap"> useful,=
 but for those that do implement this mitigation, sending the reports direc=
tly to all </span><span style=3D"font-size:10pt;font-family:Arial;color:rgb=
(0,0,0);background-color:transparent;font-style:italic;font-variant-numeric=
:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:=
pre-wrap">aggregators</span><span style=3D"font-size:10pt;font-family:Arial=
;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;=
font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap=
"> presents some meaningful challenges.</span></p><br><p dir=3D"ltr" style=
=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-=
size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;f=
ont-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:ba=
seline;white-space:pre-wrap">Finally, in terms of cost, given that the majo=
rity of the bandwidth cost is in egress cost (i.e. leaving a cloud provider=
 network.) I would agree that it would be useful to be able to specify mult=
iple locations for reports to be sent, somewhat similar to the structure pr=
oposed </span><a href=3D"https://github.com/abetterinternet/ppm-specificati=
on/issues/130#issuecomment-967730879" style=3D"text-decoration-line:none"><=
span style=3D"font-size:10pt;font-family:Arial;background-color:transparent=
;font-variant-numeric:normal;font-variant-east-asian:normal;text-decoration=
-line:underline;vertical-align:baseline;white-space:pre-wrap">in this comme=
nt</span></a><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,=
0);background-color:transparent;font-variant-numeric:normal;font-variant-ea=
st-asian:normal;vertical-align:baseline;white-space:pre-wrap"> [5]. This wo=
uld enable reports to be sent directly from clients into different cloud pr=
oviders=E2=80=93and the server running in those cloud providers could be a =
proxy of the </span><span style=3D"font-size:10pt;font-family:Arial;color:r=
gb(0,0,0);background-color:transparent;font-style:italic;font-variant-numer=
ic:normal;font-variant-east-asian:normal;vertical-align:baseline;white-spac=
e:pre-wrap">consumer </span><span style=3D"font-size:10pt;font-family:Arial=
;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;=
font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap=
">(proposed above), a proxy of the </span><span style=3D"font-size:10pt;fon=
t-family:Arial;color:rgb(0,0,0);background-color:transparent;font-style:ita=
lic;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-ali=
gn:baseline;white-space:pre-wrap">leader</span><span style=3D"font-size:10p=
t;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-vari=
ant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;w=
hite-space:pre-wrap">, or the </span><span style=3D"font-size:10pt;font-fam=
ily:Arial;color:rgb(0,0,0);background-color:transparent;font-style:italic;f=
ont-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:ba=
seline;white-space:pre-wrap">aggregators </span><span style=3D"font-size:10=
pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-var=
iant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;=
white-space:pre-wrap">themselves. This structure would allow us to separate=
 the egress cost concerns from the Sybil attack concerns.</span></p><p dir=
=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span =
style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color=
:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;ver=
tical-align:baseline;white-space:pre-wrap"><br></span></p><p style=3D"line-=
height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt=
;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-varia=
nt-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;wh=
ite-space:pre-wrap">Cheers,</span></p><p style=3D"line-height:1.38;margin-t=
op:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;c=
olor:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;fo=
nt-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">=
Erik</span></p><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;=
margin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rg=
b(0,0,0);background-color:transparent;font-variant-numeric:normal;font-vari=
ant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[1] Iss=
ue 166: adding an untrusted </span><span style=3D"font-size:10pt;font-famil=
y:Arial;color:rgb(0,0,0);background-color:transparent;font-style:italic;fon=
t-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:base=
line;white-space:pre-wrap">consumer</span><span style=3D"font-size:10pt;fon=
t-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-n=
umeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-=
space:pre-wrap"> that stores reports before aggregation: </span><a href=3D"=
https://github.com/abetterinternet/ppm-specification/issues/166" style=3D"t=
ext-decoration-line:none"><span style=3D"font-size:10pt;font-family:Arial;b=
ackground-color:transparent;font-variant-numeric:normal;font-variant-east-a=
sian:normal;text-decoration-line:underline;vertical-align:baseline;white-sp=
ace:pre-wrap">https://github.com/abetterinternet/ppm-specification/issues/1=
66</span></a></p><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;ma=
rgin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(=
0,0,0);background-color:transparent;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[2] Inter=
operable Private Attribution proposal: </span><a href=3D"https://docs.googl=
e.com/document/d/1KpdSKD8-Rn0bWPTu4UtK54ks0yv2j22pA5SrAD9av4s/edit#heading=
=3Dh.f4x9f0nqv28x" style=3D"text-decoration-line:none"><span style=3D"font-=
size:10pt;font-family:Arial;background-color:transparent;font-variant-numer=
ic:normal;font-variant-east-asian:normal;text-decoration-line:underline;ver=
tical-align:baseline;white-space:pre-wrap">https://docs.google.com/document=
/d/1KpdSKD8-Rn0bWPTu4UtK54ks0yv2j22pA5SrAD9av4s/edit#heading=3Dh.f4x9f0nqv2=
8x</span></a></p><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;ma=
rgin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(=
0,0,0);background-color:transparent;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[3] Multi=
-Browser Aggregation Service Explainer, subset of the Attribution Reporting=
 API proposal: </span><a href=3D"https://github.com/WICG/conversion-measure=
ment-api/blob/main/SERVICE.md" style=3D"text-decoration-line:none"><span st=
yle=3D"font-size:10pt;font-family:Arial;background-color:transparent;font-v=
ariant-numeric:normal;font-variant-east-asian:normal;text-decoration-line:u=
nderline;vertical-align:baseline;white-space:pre-wrap">https://github.com/W=
ICG/conversion-measurement-api/blob/main/SERVICE.md</span></a></p><p dir=3D=
"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span sty=
le=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tr=
ansparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertic=
al-align:baseline;white-space:pre-wrap">[4] Attribution Reporting API with =
Aggregate Reports, subset of the Attribution Reporting API proposal: </span=
><a href=3D"https://github.com/WICG/conversion-measurement-api/blob/main/AG=
GREGATE.md" style=3D"text-decoration-line:none"><span style=3D"font-size:10=
pt;font-family:Arial;background-color:transparent;font-variant-numeric:norm=
al;font-variant-east-asian:normal;text-decoration-line:underline;vertical-a=
lign:baseline;white-space:pre-wrap">https://github.com/WICG/conversion-meas=
urement-api/blob/main/AGGREGATE.md</span></a></p><p dir=3D"ltr" style=3D"li=
ne-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:1=
0pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-va=
riant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline=
;white-space:pre-wrap">[5] Issue 130 comment, proposing proxies to receive =
reports in separate cloud providers: </span><a href=3D"https://github.com/a=
betterinternet/ppm-specification/issues/130#issuecomment-967730879" style=
=3D"text-decoration-line:none"><span style=3D"font-size:10pt;font-family:Ar=
ial;background-color:transparent;font-variant-numeric:normal;font-variant-e=
ast-asian:normal;text-decoration-line:underline;vertical-align:baseline;whi=
te-space:pre-wrap">https://github.com/abetterinternet/ppm-specification/iss=
ues/130#issuecomment-967730879</span></a></p><br></span></div><br><div clas=
s=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 19, 202=
2 at 7:15 AM Charlie Harrison &lt;csharrison=3D<a href=3D"mailto:40google.c=
om@dmarc.ietf.org">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blo=
ckquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"=
ltr"><div>Thanks for prompting this discussion, Christopher. I would apprec=
iate your thoughts on Option 3, since it seems like there are very good rea=
sons to prefer either Option 1 or 2.</div><div><br></div><div>Another idle =
concern I had, which I cannot substantiate immediately, is whether Option #=
1 meaningfully limits the space of possible MPC protocols we could support,=
 by revealing the # of reports being sent to the helper to the leader.</div=
><div><br></div><div>cc schoppman who I think can give numbers for input sh=
ares in the current C++ implementation, as well as rkubis who likely has op=
inions here.</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" cla=
ss=3D"gmail_attr">On Tue, Jan 18, 2022 at 9:20 PM Christopher Patton &lt;cp=
atton=3D<a href=3D"mailto:40cloudflare.com@dmarc.ietf.org" target=3D"_blank=
">40cloudflare.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>Hi list,</div><div><=
/div><div></div><div><br></div><div>The current spec has the client upload =
to the leader a &quot;report&quot; that contains the encrypted input share =
of each aggregator. Later, during the aggregation flow, the leader relays t=
o each helper its share of the report (i.e., the encrypted input share, the=
 nonce, and extensions). An alternative upload flow would have the client u=
pload a report share to each aggregator directly. Both of these approaches =
have pros and cons, so in this thread I&#39;d like us to consider which app=
roach makes the most sens or whether there is a better alternative.<br></di=
v><div><br></div><div>For clarity, the current options are:<br></div><div><=
/div><div></div><div></div><div></div><div>Option #1 (status quo): Have the=
 client upload a report to the leader that contains all of the report share=
s.</div><div>Option #2 (split upload): Have the client upload a report shar=
e to each aggregator directly. For concreteness, this option is spelled out=
 in [1].</div><div><br></div><div>There are at least two considerations tha=
t lead to the current spec (i.e., option #1). One is that option #2 has mor=
e failure modes compared to option #1. For example, if the leader begins ag=
gregating a report before the helper has received its share from the client=
, then the report will be rejected and the leader will be forced to retry l=
ater on. The second consideration is that option #2 has more stringent stor=
age requirements for the helper, since it has to store its report shares fo=
r as long as they can be aggregated. (Recall that a report may be aggregate=
d multiple times, with a different=20
aggregation parameter each time [2]. This is necessary to implement=20
protocols, like Poplar [3], that require the collector to compute the=20
aggregate result before computing the next aggregation parameter. </div><di=
v><br></div><div>The major downside of option #1 (as observed in [4]) is th=
at it increases the leader-to-helper communication cost. For prio3 the over=
head is modest -- a couple hundred bytes per report share, regardless of th=
e data type -- but for poplar1 the overhead might be significant, since the=
 size of each input share is linear in the length of the input measurement.=
</div></div></blockquote><div><br></div><div>I don&#39;t quite understand t=
his. I thought in Prio the input shares were linear in the length of the in=
put measurement even excluding the proofs, since they are simple vector sha=
res. Are you talking about just the SNIP proofs?=C2=A0</div><div>=C2=A0</di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><=
br></div><div>Another downside of option #1 is that, because the leader pic=
ks the set of reports that are aggregated, it can mount a (fairly weak) Syb=
il attack in which it stuffs a batch with n-1 bogus reports, thereby leakin=
g the measurement of a target report entirely. (This requires colluding wit=
h the collector.) Taking option #2 would force the leader to upload the bog=
us report shares to the aggregators directly in order to mount the same att=
ack. In theory, there&#39;s not much difference between these two attacks, =
since we consider an attacker that controls the network. But as a practical=
 matter, it would be possible for the helper to mitigate the attack by rate=
-limiting requests from the leader&#39;s IP address. This isn&#39;t perfect=
, but it&#39;s something. (Note that devising an optional defense against S=
ybil attacks is tracked by [5].)<br></div><div><br></div><div>There are, of=
 course, alternatives to these two options. For example:<br><div>Option #3 =
(deployment specific): We could spell the protocol in a=20
way that leaves this behavior up to the deployment. In particular, for a
 given task, the servers would decide whether or not split the uploads.<br>=
</div><div>Option #4 (no re-transmission):=20
The current spec requires the leader to re-transmit the helper&#39;s report=
=20
share each time the aggregation flow is executed. Instead, we could have
 the leader transmit it once and require the helper to store it=20
thereafter.</div><div><br></div><div>Your feedback is greatly appreciated. =
If anyone has a strong preference for one of these options, sees a reason w=
hy one of them won&#39;t work, or has an idea for an alternative, please ch=
ime in :)</div><div><br></div><div> Also, it would be great if someone who =
has implemented IDPFs could tell us how large the input shares are for, say=
, inputs of length 64, 128, and 256 bits.</div><div><br></div><div>Best, <b=
r></div><div>Chris P.<br></div></div><br><div><br></div><div>[1] PR: Draft =
for split upload: <a href=3D"https://github.com/abetterinternet/ppm-specifi=
cation/pull/174" target=3D"_blank">https://github.com/abetterinternet/ppm-s=
pecification/pull/174</a></div><div></div><div>[2] Section of VDAF spec def=
ining the aggregation parameter: <a href=3D"https://cjpatton.github.io/vdaf=
/draft-patton-cfrg-vdaf.html#name-preparation" target=3D"_blank">https://cj=
patton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-preparation</a></div=
><div>[3] Section of VDAF spec describing Poplar: <a href=3D"https://cjpatt=
on.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1" target=3D"_blan=
k">https://cjpatton.github.io/vdaf/draft-patton-cfrg-vdaf.html#name-poplar1=
</a></div><div></div><div>[4] Issue: Cost implications of the leader being =
the only upload endpoint: <a href=3D"https://github.com/abetterinternet/ppm=
-specification/issues/130" target=3D"_blank">https://github.com/abetterinte=
rnet/ppm-specification/issues/130</a></div><div>[5] Issue: Client attestati=
on: <a href=3D"https://github.com/abetterinternet/ppm-specification/issues/=
89" target=3D"_blank">https://github.com/abetterinternet/ppm-specification/=
issues/89</a></div><div><br></div></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--000000000000b8890905d5f251e1--


From nobody Wed Jan 19 12:59:46 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id ECF8C3A1C67 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 12:59:38 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id CCH5TImSK3js for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 12:59:34 -0800 (PST)
Received: from mail-pg1-x532.google.com (mail-pg1-x532.google.com [IPv6:2607:f8b0:4864:20::532]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 0EE523A1C07 for <ppm@ietf.org>; Wed, 19 Jan 2022 12:59:27 -0800 (PST)
Received: by mail-pg1-x532.google.com with SMTP id p125so3745248pga.2 for <ppm@ietf.org>; Wed, 19 Jan 2022 12:59:27 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=JLKDx9J5CD3Ndd4e6DqUFIcnpi3AksKJdnTlXadt0zs=; b=bsyOd5jL7Ev4nnkitF1ve5Mtos6RTvZnUyAqbTpxf4/wWSldIjbDBR1pUk8IghMPMn C+5/RwCNBJXCgybUXy6iLUwdCQj/zheaI8cSU6S1gZsUoydg1JhGK2Oj1jhVuXGrdTsL NimwgN+Oak9ND3MWQpi4rMXCVJOaRj5zprSbI=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=JLKDx9J5CD3Ndd4e6DqUFIcnpi3AksKJdnTlXadt0zs=; b=joALyvej/5PokaoO8DknkQcJpjyaNPrqgkt4+rpaD9NCvZUltGhaFAMlVDR0VkyH/j wtiOf9h7EntLhJBuMUtyRs2SqtJUzkYXdyM5+ilkYPevDwN9OOkcxOC9ENFM+RXtKTBF Z1vAp3gCNc3ux6ePK+JfrD7SIVXv2CTWfPvFS/YDjVbh51LnV8oKLNij7dNPRSCvUtFp SHuhmPvosgs6Kxl24vRICcErhsuuxxsY65ou70aFcAWGSSFxg5NEHsJQCnOfQ4PltRVu 31ccer5oEOUEhJww6z4yE0AVKpdXWNw5446imMN/QAQxarnSln4i2ixAY/xHhGuagDX7 8zfA==
X-Gm-Message-State: AOAM532dV2s1dt9VZ7zZ+9nw+BFz4IOh7rPYau9Rh6f31w9qN0CjgGwa snJ+95VBXAUSb4Xn6lBoo8IrO+AIwL4+rqRLsf9CdA==
X-Google-Smtp-Source: ABdhPJwLWp55ptJyX2J4ZfS4HEx3b3X9lF4yv1rfnp8DIJNIU0LxI/XKmdmiQZpSh2ofUumeSO0e8W7EJ7qNdeP3YS4=
X-Received: by 2002:a63:cf13:: with SMTP id j19mr28191110pgg.279.1642625966626;  Wed, 19 Jan 2022 12:59:26 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com>
In-Reply-To: <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Wed, 19 Jan 2022 12:59:15 -0800
Message-ID: <CAG2Zi21bFN30J-MsrURSqsgS3Azru47FEPpRQ=7HoQ35K265uw@mail.gmail.com>
To: Charlie Harrison <csharrison=40google.com@dmarc.ietf.org>
Cc: ppm@ietf.org, Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>
Content-Type: multipart/alternative; boundary="0000000000003d912c05d5f5a966"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/706k2MsMQRxrGDTmRUhfvH7FdY4>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 20:59:45 -0000

--0000000000003d912c05d5f5a966
Content-Type: text/plain; charset="UTF-8"

> I would appreciate your thoughts on Option 3, since it seems like there
> are very good reasons to prefer either Option 1 or 2.
>

I think the main downside to #3 is that it adds complexity to the protocol
and therefore more ways for things to fail. Ideally we'd identify a single
approach that everyone agrees will work and then figure out how to mitigate
the downsides.  In my opinion, option #3 should only be considered if there
is some use case for which #1 or #2 is untenable.



> Another idle concern I had, which I cannot substantiate immediately, is
> whether Option #1 meaningfully limits the space of possible MPC protocols
> we could support, by revealing the # of reports being sent to the helper to
> the leader.
>

This is an interesting observation. The current PPM spec is centered around
the execution of a VDAF, the shape of which already rules out any MPC in
which the helper and leader get a different number of shares. This could
change, of course, as we're amendable to changes to the VDAF spec to
include MPC protocols that might end up being useful. Or there might be
different PPM variants, in order to support MPC protocols with a different
shape.



> The major downside of option #1 (as observed in [4]) is that it increases
>> the leader-to-helper communication cost. For prio3 the overhead is modest
>> -- a couple hundred bytes per report share, regardless of the data type --
>> but for poplar1 the overhead might be significant, since the size of each
>> input share is linear in the length of the input measurement.
>>
>
> I don't quite understand this. I thought in Prio the input shares were
> linear in the length of the input measurement even excluding the proofs,
> since they are simple vector shares. Are you talking about just the SNIP
> proofs?
>

Yes, this includes the proof shares. In prio3, the input and proof shares
are just vectors of field elements that add up to the plaintext input and
proof. The helper shares are computed by picking a PRNG seed and expanding
it into a sequence of pseudorandom field elements. That means all a helper
needs to represent its share is the seed itself. This optimization is
fairly standard. Here is how it works in code:
https://github.com/abetterinternet/libprio-rs/blob/main/src/vdaf/prio3.rs#L417-L441

Note that the leader's input share is indeed linear in the length of the
measurement. The length of the proof share depends on the data type, but
can range anywhere from O(n), O(n^{-1/2}), or O(log n) with some
optimization.

Chris P.

--0000000000003d912c05d5f5a966
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmai=
l_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,20=
4,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"><div>I would app=
reciate your thoughts on Option 3, since it seems like there are very good =
reasons to prefer either Option 1 or 2.</div></div></div></blockquote><div>=
<br></div><div>I think the main downside to #3 is that it adds complexity t=
o the protocol and therefore more ways for things to fail. Ideally we&#39;d=
 identify a single approach that everyone agrees will work and then figure =
out how to mitigate the downsides.=C2=A0 In my opinion, option #3 should on=
ly be considered if there is some use case for which #1 or #2 is untenable.=
</div><br><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"margi=
n:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex=
"><div dir=3D"ltr"><div dir=3D"ltr"><div>Another idle concern I had, which =
I cannot substantiate immediately, is whether Option #1 meaningfully limits=
 the space of possible MPC protocols we could support, by revealing the # o=
f reports being sent to the helper to the leader.</div></div></div></blockq=
uote><div><br></div><div>This is an interesting observation. The current PP=
M spec is centered around the execution of a VDAF, the shape of which alrea=
dy rules out any MPC in which the helper and leader get a different number =
of shares. This could change, of course, as we&#39;re amendable to changes =
to the VDAF spec to include MPC protocols that might end up being useful. O=
r there might be different PPM variants, in order to support MPC protocols =
with a different shape.</div><div>=C2=A0</div><div>=C2=A0</div><blockquote =
class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px sol=
id rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div class=3D"gmail_=
quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;=
border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><=
div>The major downside of option #1 (as observed in [4]) is that it increas=
es the leader-to-helper communication cost. For prio3 the overhead is modes=
t -- a couple hundred bytes per report share, regardless of the data type -=
- but for poplar1 the overhead might be significant, since the size of each=
 input share is linear in the length of the input measurement.</div></div><=
/blockquote><div><br></div><div>I don&#39;t quite understand this. I though=
t in Prio the input shares were linear in the length of the input measureme=
nt even excluding the proofs, since they are simple vector shares. Are you =
talking about just the SNIP proofs?</div></div></div></blockquote><div><br>=
</div><div>Yes, this includes the proof shares. In prio3, the input and pro=
of shares are just vectors of field elements that add up to the plaintext i=
nput and proof. The helper shares are computed by picking a PRNG seed and e=
xpanding it into a sequence of pseudorandom field elements. That means all =
a helper needs to represent its share is the seed itself. This optimization=
 is fairly standard. Here is how it works in code: <a href=3D"https://githu=
b.com/abetterinternet/libprio-rs/blob/main/src/vdaf/prio3.rs#L417-L441">htt=
ps://github.com/abetterinternet/libprio-rs/blob/main/src/vdaf/prio3.rs#L417=
-L441</a></div><div><br></div><div>Note that the leader&#39;s input share i=
s indeed linear in the length of the measurement. The length of the proof s=
hare depends on the data type, but can range anywhere from O(n), O(n^{-1/2}=
), or O(log n) with some optimization.<br></div></div><div class=3D"gmail_q=
uote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>

--0000000000003d912c05d5f5a966--


From nobody Wed Jan 19 13:25:25 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 16F8A3A1C7D for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 13:25:24 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: 0.611
X-Spam-Level: 
X-Spam-Status: No, score=0.611 tagged_above=-999 required=5 tests=[BAYES_50=0.8, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_KAM_HTML_FONT_INVALID=0.01, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id CCyJmuJfSa1P for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 13:25:19 -0800 (PST)
Received: from mail-pl1-x633.google.com (mail-pl1-x633.google.com [IPv6:2607:f8b0:4864:20::633]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id AD8453A1C7A for <ppm@ietf.org>; Wed, 19 Jan 2022 13:25:19 -0800 (PST)
Received: by mail-pl1-x633.google.com with SMTP id c3so3415074pls.5 for <ppm@ietf.org>; Wed, 19 Jan 2022 13:25:19 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=wwzKqpwjySvaGBxqUCan2RSB/0wqg3ClO4UxEXAH48I=; b=lPnW0yhj4gVPPcPyWu7SpdPWDZS2tPUs4LCWV3ldLjYo5xYw8VxiALG3EAwruMZliD aRrjX1Dghd9y4VU1x76Xc0rXcnnfHHPKpTi4kXzoJiX88gfGweg8GS6iJiJXTjQw7m2d M1XOfBnBZ/eDE0iuLDbdZeZCnRYdYCX4pSvDA=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=wwzKqpwjySvaGBxqUCan2RSB/0wqg3ClO4UxEXAH48I=; b=CJBKZGvW59imaD95dDVjzKxd9AGsIMR8TRgH0qM9z4y99Z6+fVkMpbNttKYewBtg6h YE1jh3BmwfcU53yUxWJ6wwkPqjDSmhhwf5Uj7ZWN11TaDC0LB3PDU9DKvr3KlodjscKC 3ABw916ts69/e8UZHZ6Pw1KjkidQt81xzkPMpBUU+R6/6FLG8U5fl+g4LcBz44gD/H10 4MbfXo4GE0NXAMdWN00hPVkN/cHHq8kkkPxJ40hLDA1IaZk4zDb0hqi9t2Wt+2L/RPX0 pbWapoNUyjCf2zumcq+I7aOLIUaM9Kh+ZdHfX0f50O1rq1k+65mS3wrQYnnmccl8gG8Z dwtw==
X-Gm-Message-State: AOAM532Uc5hkv40PyrqlKI8qVRwLqRsjvMHGzjg9Str4Oeesa+OcQzO5 CxZ/MvUgVIpOgOcctAKPnIRPTrlkaDXwESLOrL5rXQ==
X-Google-Smtp-Source: ABdhPJyCRoYmiG80hUozzkyq04YW6kqfEWnIPSJnsPQI18OtVaD6pIl/LyTJhCj9gaSIK07UiMu0XTsmWDHvHBeosPc=
X-Received: by 2002:a17:903:2308:b0:14a:f22b:f604 with SMTP id d8-20020a170903230800b0014af22bf604mr4344164plh.48.1642627517006; Wed, 19 Jan 2022 13:25:17 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com>
In-Reply-To: <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Wed, 19 Jan 2022 13:25:06 -0800
Message-ID: <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Charlie Harrison <csharrison@google.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000a66b5705d5f605bb"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/6cwfkk-0eEqNSD-vvzQBU0Zih5o>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 21:25:24 -0000

--000000000000a66b5705d5f605bb
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi Erik!


> There is a 4th option being discussed in issue #166
> <https://github.com/abetterinternet/ppm-specification/issues/166> [1],
> which moves in the opposite direction by proposing sending reports to a n=
ew
> party called a consumer which is entirely untrusted and specified by the
> collector. Note that the consumer would also also be able to mount the
> Sybil attack, which I=E2=80=99ll address below.
>

Let's call this option #5 for the sake of discussion. Option #4 is defined
at the top of this thread.


> The rationale here is almost entirely operational. Both of the APIs above
> are implemented within an user-agent, but are used by arbitrary websites
> and apps. Consuming and storing these reports sent from instances of thos=
e
> user-agents is a non-trivial infrastructure challenge, and we should assu=
me
> that aggregators will not have 100% uptime. In fact, improving from 99.9%
> uptime to 99.99% uptime can often mean an order of magnitude increase in
> operational costs.
>
> I would assume that the set of parties offering their services as
> aggregators will be fairly constrained for the near-to-mid term, and as
> such, collectors will have minimal choice when it comes to finding the
> appropriate trade offs between operational concerns like uptime vs
> infrastructure costs. As such, I believe it=E2=80=99s beneficial to move =
as much of
> this cost and operational effort outside of the aggregator
>

The ENPA system called this the "ingestor". There the rationale was the
same: To relieve the aggregators of the operational burden of horizontally
scaling to consume client traffic. In the current spec, there is a
distinguished aggregator, called the leader, who bears this responsibility.

I think it makes sense for a leader deployment to have a separate
"leader-ingestor" service and "leader-aggregator" service that have
different uptime requirements (and are perhaps maintained by different
teams within the same organization). My (current) objection to option #5 is
that roles like "leader", "helper", and "collector" have more to do with
the trust model than who is operationally responsible for what. For
instance, a statement "We assume the leader and helper don't collude" is
meaningful because they get different sets of input shares; on the other
hand, it doesn't sense to ask whether the leader-ingestor and
leader-aggregator collude, since they only see one set of input shares.

It's just my opinion, but I think option #5 would only make sense if the
ingestor/consumer did something that is not already implemented by some
server in the protocol.


> To the point on Sybil attacks: nearly all the APIs that would implement
> PPM will run in user-agents on untrusted and unknown devices. The code us=
ed
> to generate those reports will be open source, or at least the protocol
> will be publicly known, so the bigger threat, in my opinion, is a malicio=
us
> entity operating a modified user-agent to generate events which enable a
> Sybil attack. The IPA proposal linked above (as well as the Attribution
> Reporting API with Aggregate Reports
> <https://github.com/WICG/conversion-measurement-api/blob/main/AGGREGATE.m=
d>
> [4]) propose using differential privacy to address these types of attacks=
.
> This approach should also mitigate Sybil attacks.
>

DP has been discussed as a defense for Sybil attacks, as has (some variant
of) Privacy Pass, which would allow the aggregators to avoid ingesting
reports from clients they haven't seen before. All of these things are
still on the table, but so far have been considered as extensions to the
core PPM protocol.



> I don=E2=80=99t believe that all implementations will end up using differ=
ential
> privacy, and thus may find the proposal for reports to be sent directly t=
o
> all aggregators useful, but for those that do implement this mitigation,
> sending the reports directly to all aggregators presents some meaningful
> challenges.
>

Interesting, could you be more specific? I'm not aware of a DP mechanism
for which option #2 would be a barrier.

--000000000000a66b5705d5f605bb
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div class=3D"gmail_quote"><div>Hi Erik!<br></div><div>=C2=
=A0</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8e=
x;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"=
><span id=3D"m_1487501867054257734gmail-m_2460893677559864241gmail-docs-int=
ernal-guid-7091e80b-7fff-7c0e-b2b6-16a440521498"><p dir=3D"ltr" style=3D"li=
ne-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:1=
0pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-va=
riant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline=
;white-space:pre-wrap">There is a 4th option being discussed in </span><a h=
ref=3D"https://github.com/abetterinternet/ppm-specification/issues/166" sty=
le=3D"text-decoration-line:none" target=3D"_blank"><span style=3D"font-size=
:10pt;font-family:Arial;background-color:transparent;font-variant-numeric:n=
ormal;font-variant-east-asian:normal;text-decoration-line:underline;vertica=
l-align:baseline;white-space:pre-wrap">issue #166</span></a><span style=3D"=
font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transpar=
ent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-ali=
gn:baseline;white-space:pre-wrap"> [1], which moves in the opposite directi=
on by proposing sending reports to a new party called a </span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-style:italic;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">consumer</span><spa=
n style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-col=
or:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;v=
ertical-align:baseline;white-space:pre-wrap"> which is entirely untrusted a=
nd specified by the </span><span style=3D"font-size:10pt;font-family:Arial;=
color:rgb(0,0,0);background-color:transparent;font-style:italic;font-varian=
t-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;whi=
te-space:pre-wrap">collector</span><span style=3D"font-size:10pt;font-famil=
y:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:=
normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:p=
re-wrap">. Note that the </span><span style=3D"font-size:10pt;font-family:A=
rial;color:rgb(0,0,0);background-color:transparent;font-style:italic;font-v=
ariant-numeric:normal;font-variant-east-asian:normal;vertical-align:baselin=
e;white-space:pre-wrap">consumer</span><span style=3D"font-size:10pt;font-f=
amily:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-nume=
ric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-spa=
ce:pre-wrap"> would also also be able to mount the Sybil attack, which I=E2=
=80=99ll address below.</span></p></span></div></blockquote><div><br></div>=
<div>Let&#39;s call this option #5 for the sake of discussion. Option #4 is=
 defined at the top of this thread.<br></div>=C2=A0<br><blockquote class=3D=
"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(2=
04,204,204);padding-left:1ex"><div><span id=3D"m_1487501867054257734gmail-m=
_2460893677559864241gmail-docs-internal-guid-7091e80b-7fff-7c0e-b2b6-16a440=
521498"><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bott=
om:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);ba=
ckground-color:transparent;font-variant-numeric:normal;font-variant-east-as=
ian:normal;vertical-align:baseline;white-space:pre-wrap">The rationale here=
 is almost entirely operational. Both of the APIs above are implemented wit=
hin an user-agent, but are used by arbitrary websites and apps. Consuming a=
nd storing these reports sent from instances of those user-agents is a non-=
trivial infrastructure challenge, and we should assume that </span><span st=
yle=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:t=
ransparent;font-style:italic;font-variant-numeric:normal;font-variant-east-=
asian:normal;vertical-align:baseline;white-space:pre-wrap">aggregators</spa=
n><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);backgrou=
nd-color:transparent;font-variant-numeric:normal;font-variant-east-asian:no=
rmal;vertical-align:baseline;white-space:pre-wrap"> will not have 100% upti=
me. In fact, improving from 99.9% uptime to 99.99% uptime can often mean an=
 order of magnitude increase in operational costs.</span></p><br><p dir=3D"=
ltr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span styl=
e=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tra=
nsparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertica=
l-align:baseline;white-space:pre-wrap">I would assume that the set of parti=
es offering their services as </span><span style=3D"font-size:10pt;font-fam=
ily:Arial;color:rgb(0,0,0);background-color:transparent;font-style:italic;f=
ont-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:ba=
seline;white-space:pre-wrap">aggregators</span><span style=3D"font-size:10p=
t;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-vari=
ant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;w=
hite-space:pre-wrap"> will be fairly constrained for the near-to-mid term, =
and as such, </span><span style=3D"font-size:10pt;font-family:Arial;color:r=
gb(0,0,0);background-color:transparent;font-style:italic;font-variant-numer=
ic:normal;font-variant-east-asian:normal;vertical-align:baseline;white-spac=
e:pre-wrap">collectors</span><span style=3D"font-size:10pt;font-family:Aria=
l;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal=
;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wra=
p"> will have minimal choice when it comes to finding the appropriate trade=
 offs between operational concerns like uptime vs infrastructure costs. As =
such, I believe it=E2=80=99s beneficial to move as much of this cost and op=
erational effort outside of the </span><span style=3D"font-size:10pt;font-f=
amily:Arial;color:rgb(0,0,0);background-color:transparent;font-style:italic=
;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:=
baseline;white-space:pre-wrap">aggregator</span></p></span></div></blockquo=
te><div><br></div><div>The ENPA system called this the &quot;ingestor&quot;=
. There the rationale was the same: To relieve the aggregators of the opera=
tional burden of horizontally scaling to consume client traffic. In the cur=
rent spec, there is a distinguished aggregator, called the leader, who bear=
s this responsibility.</div><div><br></div>I think it makes sense for a lea=
der deployment to have a separate &quot;leader-ingestor&quot; service and &=
quot;leader-aggregator&quot; service that have different uptime requirement=
s (and are perhaps maintained by different teams within the same organizati=
on). My (current) objection to option #5 is that roles like &quot;leader&qu=
ot;, &quot;helper&quot;, and &quot;collector&quot; have more to do with the=
 trust model than who is operationally responsible for what. For instance, =
a statement &quot;We assume the leader and helper don&#39;t collude&quot; i=
s meaningful because they get different sets of input shares; on the other =
hand, it doesn&#39;t sense to ask whether the leader-ingestor and leader-ag=
gregator collude, since they only see one set of input shares.</div><div cl=
ass=3D"gmail_quote"><br></div><div class=3D"gmail_quote">It&#39;s just my o=
pinion, but I think option #5 would only make sense if the ingestor/consume=
r did something that is not already implemented by some server in the proto=
col.<br></div><div class=3D"gmail_quote"><div>=C2=A0</div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><span id=3D"m_14875018670=
54257734gmail-m_2460893677559864241gmail-docs-internal-guid-7091e80b-7fff-7=
c0e-b2b6-16a440521498"><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:=
0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;colo=
r:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-=
variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">To =
the point on Sybil attacks: nearly all the APIs that would implement PPM wi=
ll run in user-agents on untrusted and unknown devices. The code used to ge=
nerate those reports will be open source, or at least the protocol will be =
publicly known, so the bigger threat, in my opinion, is a malicious entity =
operating a modified user-agent to generate events which enable a Sybil att=
ack. The IPA proposal linked above (as well as the </span><a href=3D"https:=
//github.com/WICG/conversion-measurement-api/blob/main/AGGREGATE.md" style=
=3D"text-decoration-line:none" target=3D"_blank"><span style=3D"font-size:1=
0pt;font-family:Arial;background-color:transparent;font-variant-numeric:nor=
mal;font-variant-east-asian:normal;text-decoration-line:underline;vertical-=
align:baseline;white-space:pre-wrap">Attribution Reporting API with Aggrega=
te Reports</span></a><span style=3D"font-size:10pt;font-family:Arial;color:=
rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-va=
riant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap"> [4])=
 propose using differential privacy to address these types of attacks. This=
 approach should also mitigate Sybil attacks.</span></p></span></div></bloc=
kquote><div><br></div><div>DP has been discussed as a defense for Sybil att=
acks, as has (some variant of) Privacy Pass, which would allow the aggregat=
ors to avoid ingesting reports from clients they haven&#39;t seen before. A=
ll of these things are still on the table, but so far have been considered =
as extensions to the core PPM protocol.<br></div><div><br></div><div>=C2=A0=
</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;b=
order-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><s=
pan id=3D"m_1487501867054257734gmail-m_2460893677559864241gmail-docs-intern=
al-guid-7091e80b-7fff-7c0e-b2b6-16a440521498"><p dir=3D"ltr" style=3D"line-=
height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt=
;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-varia=
nt-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;wh=
ite-space:pre-wrap">I don=E2=80=99t believe that all implementations will e=
nd up using differential privacy, and thus may find the proposal for report=
s to be sent directly to all </span><span style=3D"font-size:10pt;font-fami=
ly:Arial;color:rgb(0,0,0);background-color:transparent;font-style:italic;fo=
nt-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:bas=
eline;white-space:pre-wrap">aggregators</span><span style=3D"font-size:10pt=
;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-varia=
nt-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;wh=
ite-space:pre-wrap"> useful, but for those that do implement this mitigatio=
n, sending the reports directly to all </span><span style=3D"font-size:10pt=
;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-style=
:italic;font-variant-numeric:normal;font-variant-east-asian:normal;vertical=
-align:baseline;white-space:pre-wrap">aggregators</span><span style=3D"font=
-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;=
font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:b=
aseline;white-space:pre-wrap"> presents some meaningful challenges.</span><=
/p></span></div></blockquote><div><br></div><div>Interesting, could you be =
more specific? I&#39;m not aware of a DP mechanism for which option #2 woul=
d be a barrier.<br></div></div></div>

--000000000000a66b5705d5f605bb--


From nobody Wed Jan 19 14:31:33 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 8E7873A1F7B for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 14:31:31 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.086
X-Spam-Level: 
X-Spam-Status: No, score=-2.086 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_KAM_HTML_FONT_INVALID=0.01, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id NN2OresMvYC5 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 14:31:26 -0800 (PST)
Received: from mail-lf1-x12c.google.com (mail-lf1-x12c.google.com [IPv6:2a00:1450:4864:20::12c]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 41ABA3A1F78 for <ppm@ietf.org>; Wed, 19 Jan 2022 14:31:26 -0800 (PST)
Received: by mail-lf1-x12c.google.com with SMTP id o15so14004771lfo.11 for <ppm@ietf.org>; Wed, 19 Jan 2022 14:31:26 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=SFM8vgLNZ9L81pdVe1EOgicY66EKP0KeeONrfq0zJss=; b=ciMDInaWlC5JDFR/SNe016Tsy8SKzi6WZ3VumcOtQo8dPBZ/y6G+Qkau32c+md73Yu FTxDNeLmnEHs+dmVhXsFRUaUJFzyJSzCu7OuIeyPcnLqrKB5W5MS6M63ZwSiqLq+sBuQ rZtZZVIMQZe3DZT9gxd68IOMwIaKrIu2ElkDuF7SmHs+cFJEeYn7D5G25DSBA0g1lKuX sL2nrYr4jh9WLCINSZmIZ1HxczA2hb4W/VRCfIw2jd4HNQQMmVzlI7Bt1g0I3dV7IYoW n+cPDJOmFVRb2m2E9NIVXw92wBAwIv8IwEcWprAGvGQazsa/P3sOX2Xlv3owzk0IHq4I O7Dw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=SFM8vgLNZ9L81pdVe1EOgicY66EKP0KeeONrfq0zJss=; b=CQDIy3ofsvKxBVrrfeZ55tEwIP0WBaiEEb0ZKvyXFA+a1U6/Y8LKB3Ohy+5wPtoQ4G DrZa9ASnJJfC9vLl2UNWSJ1IrnQvAK/zQIt0AP9uuAiY7GRzPB5qv4kZ94S9S34uTzy2 VaowRtWo6jjYpfNMWsFgF8KAm39g0oC59rAPvPTWW8nOptoTNJ9Nog6QWMdh59fRev2Q TLpvcAhdyu+3i+mU8EQ2ETCGQQpZ3BURB9Xnub/Ar86dQaGRJsv30w1F2obN5CeP8NuW BGQR6jBLL0oCELldp8vL9GNTjXRbulwyEFekKFvSQYvzoa9iYuORxri7/6PoCdQwSuFA CXIA==
X-Gm-Message-State: AOAM532AK5dM2+1tat85c2xFk5lzBOQ/mzaiQsOr8PRyMJ3ShnZKUnOZ +3FtGUtpUqn8uW/LcYQk/OUdBNPHlBeCrB8y1eE=
X-Google-Smtp-Source: ABdhPJxXHUdM10ZHV1NQr2FD6vvMpqI3JDAZ3jq6NsKnT6nQJXYvSpl8C4RHX1M3svsz1MLSwRsPdMwin5WG3Gn6inI=
X-Received: by 2002:a05:6512:16a6:: with SMTP id bu38mr30288912lfb.168.1642631482190;  Wed, 19 Jan 2022 14:31:22 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com>
In-Reply-To: <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Wed, 19 Jan 2022 14:31:10 -0800
Message-ID: <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com>
To: Christopher Patton <cpatton@cloudflare.com>
Cc: Charlie Harrison <csharrison@google.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000fe2f0205d5f6f123"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/9aqlul8oRYGfUtrTRH0kDg5he4I>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 22:31:32 -0000

--000000000000fe2f0205d5f6f123
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Thanks for the reply Chris.

I think it makes sense for a leader deployment to have a separate
> "leader-ingestor" service and "leader-aggregator" service that have
> different uptime requirements (and are perhaps maintained by different
> teams within the same organization).


What I am proposing, and is discussed in issue #166 and #64, is allowing an
entirely untrusted entity to play the role of "ingestor", so that a
"collector" can either run this system themselves, or be able to hire any
service (beyond the set of approved organizations that run aggregators.)
This is particularly important in the setting where a web platform vendor
may have a very small set of approved organizations; shifting as much of
the operational cost out of that small set.

>From the "collector" point of view, I see these options as
Option #1: I have a limited set of leaders to choose from which can receive
my reports, without much optionality between uptime and cost.
Option #2: I have an even more limited set of aggregators which must
receive my reports, and uptime is dependent on all of them.
Option #5: I can choose to ingest reports myself, or outsource it to any
service provider. I still have a limited set of aggregators, but I only
rely on them for aggregation and not real-time ingestion.

If we can design solutions in which option #5 still provides the intended
security and privacy properties, it seems greatly preferable to the
"collector". That said, I am advocating for option #3, with the choice that
is inclusive of option #5.


> My (current) objection to option #5 is that roles like "leader", "helper"=
,
> and "collector" have more to do with the trust model than who is
> operationally responsible for what. ... it doesn't sense to ask whether t=
he
> leader-ingestor and leader-aggregator collude, since they only see one se=
t
> of input shares.


There is a meaningful set of attacks where the "ingestor" is or colludes
with the "collector", given that the collector may have added context. For
example, in this case the "collector" and "ingestor" could work together to
mount the Sybil attack you describe.

I don=E2=80=99t believe that all implementations will end up using differen=
tial
> privacy, and thus may find the proposal for reports to be sent directly t=
o
> all aggregators useful, but for those that do implement this mitigation,
> sending the reports directly to all aggregators presents some meaningful
> challenges.
> Interesting, could you be more specific? I'm not aware of a DP mechanism
> for which option #2 would be a barrier.


I'm not suggesting that option #2 would prevent DP, but rather that a
solution which employs a DP mechanism to prevent the Sybil attack incurs
the added operational costs of option #2, without the added benefit.

Cheers,
Erik

On Wed, Jan 19, 2022 at 1:25 PM Christopher Patton <cpatton@cloudflare.com>
wrote:

> Hi Erik!
>
>
>> There is a 4th option being discussed in issue #166
>> <https://github.com/abetterinternet/ppm-specification/issues/166> [1],
>> which moves in the opposite direction by proposing sending reports to a =
new
>> party called a consumer which is entirely untrusted and specified by the
>> collector. Note that the consumer would also also be able to mount the
>> Sybil attack, which I=E2=80=99ll address below.
>>
>
> Let's call this option #5 for the sake of discussion. Option #4 is define=
d
> at the top of this thread.
>
>
>> The rationale here is almost entirely operational. Both of the APIs abov=
e
>> are implemented within an user-agent, but are used by arbitrary websites
>> and apps. Consuming and storing these reports sent from instances of tho=
se
>> user-agents is a non-trivial infrastructure challenge, and we should ass=
ume
>> that aggregators will not have 100% uptime. In fact, improving from
>> 99.9% uptime to 99.99% uptime can often mean an order of magnitude incre=
ase
>> in operational costs.
>>
>> I would assume that the set of parties offering their services as
>> aggregators will be fairly constrained for the near-to-mid term, and as
>> such, collectors will have minimal choice when it comes to finding the
>> appropriate trade offs between operational concerns like uptime vs
>> infrastructure costs. As such, I believe it=E2=80=99s beneficial to move=
 as much of
>> this cost and operational effort outside of the aggregator
>>
>
> The ENPA system called this the "ingestor". There the rationale was the
> same: To relieve the aggregators of the operational burden of horizontall=
y
> scaling to consume client traffic. In the current spec, there is a
> distinguished aggregator, called the leader, who bears this responsibilit=
y.
>
> I think it makes sense for a leader deployment to have a separate
> "leader-ingestor" service and "leader-aggregator" service that have
> different uptime requirements (and are perhaps maintained by different
> teams within the same organization). My (current) objection to option #5 =
is
> that roles like "leader", "helper", and "collector" have more to do with
> the trust model than who is operationally responsible for what. For
> instance, a statement "We assume the leader and helper don't collude" is
> meaningful because they get different sets of input shares; on the other
> hand, it doesn't sense to ask whether the leader-ingestor and
> leader-aggregator collude, since they only see one set of input shares.
>
> It's just my opinion, but I think option #5 would only make sense if the
> ingestor/consumer did something that is not already implemented by some
> server in the protocol.
>
>
>> To the point on Sybil attacks: nearly all the APIs that would implement
>> PPM will run in user-agents on untrusted and unknown devices. The code u=
sed
>> to generate those reports will be open source, or at least the protocol
>> will be publicly known, so the bigger threat, in my opinion, is a malici=
ous
>> entity operating a modified user-agent to generate events which enable a
>> Sybil attack. The IPA proposal linked above (as well as the Attribution
>> Reporting API with Aggregate Reports
>> <https://github.com/WICG/conversion-measurement-api/blob/main/AGGREGATE.=
md>
>> [4]) propose using differential privacy to address these types of attack=
s.
>> This approach should also mitigate Sybil attacks.
>>
>
> DP has been discussed as a defense for Sybil attacks, as has (some varian=
t
> of) Privacy Pass, which would allow the aggregators to avoid ingesting
> reports from clients they haven't seen before. All of these things are
> still on the table, but so far have been considered as extensions to the
> core PPM protocol.
>
>
>
>> I don=E2=80=99t believe that all implementations will end up using diffe=
rential
>> privacy, and thus may find the proposal for reports to be sent directly =
to
>> all aggregators useful, but for those that do implement this mitigation,
>> sending the reports directly to all aggregators presents some meaningful
>> challenges.
>>
>
> Interesting, could you be more specific? I'm not aware of a DP mechanism
> for which option #2 would be a barrier.
>

--000000000000fe2f0205d5f6f123
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Thanks for the reply Chris.<div><br></div><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left-width:1px;b=
order-left-style:solid;border-left-color:rgb(204,204,204);padding-left:1ex"=
>I think it makes sense for a leader deployment to have a separate &quot;le=
ader-ingestor&quot; service and &quot;leader-aggregator&quot; service that =
have different uptime requirements (and are perhaps maintained by different=
 teams within the same organization).</blockquote><div><br></div><div>What =
I am proposing, and is discussed=C2=A0in issue #166 and #64, is allowing an=
 entirely untrusted entity to play the role of &quot;ingestor&quot;, so tha=
t a &quot;collector&quot; can either run this system themselves, or be able=
 to hire any service (beyond the set of approved organizations that run agg=
regators.) This is particularly important in the setting where a web platfo=
rm vendor may have a very small set of approved organizations; shifting as =
much of the operational cost out of that small set.</div><div><br></div><di=
v>From the &quot;collector&quot; point of view, I see these options as</div=
><div>Option #1: I have a limited set of leaders to choose from which can r=
eceive my reports, without much optionality between uptime and cost.</div><=
div>Option #2: I have an even more limited set of aggregators which must re=
ceive my reports, and uptime is dependent on all of them.</div><div>Option =
#5: I can choose to ingest reports myself, or outsource it to any service p=
rovider. I still have a limited set of aggregators, but I only rely on them=
 for aggregation and not real-time ingestion.</div><div><br></div><div>If w=
e can design solutions in which option #5 still provides the intended secur=
ity and privacy properties, it seems greatly preferable to the &quot;collec=
tor&quot;. That said, I am advocating for option #3, with the choice that i=
s inclusive of option #5.</div><div>=C2=A0</div><blockquote class=3D"gmail_=
quote" style=3D"margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-=
style:solid;border-left-color:rgb(204,204,204);padding-left:1ex">My (curren=
t) objection to option #5 is that roles like &quot;leader&quot;, &quot;help=
er&quot;, and &quot;collector&quot; have more to do with the trust model th=
an who is operationally responsible for what. ... it doesn&#39;t sense to a=
sk whether the leader-ingestor and leader-aggregator collude, since they on=
ly see one set of input shares.</blockquote><div><br>There is a meaningful =
set of attacks where the &quot;ingestor&quot; is or colludes with the &quot=
;collector&quot;, given that the collector may have added context. For exam=
ple, in this case the &quot;collector&quot; and &quot;ingestor&quot; could =
work together to mount the Sybil attack you describe.</div><div><br></div><=
blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-l=
eft-width:1px;border-left-style:solid;border-left-color:rgb(204,204,204);pa=
dding-left:1ex"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0=
,0,0);font-variant-numeric:normal;font-variant-east-asian:normal;vertical-a=
lign:baseline;white-space:pre-wrap">I don=E2=80=99t believe that all implem=
entations will end up using differential privacy, and thus may find the pro=
posal for reports to be sent directly to all </span><span style=3D"font-siz=
e:10pt;font-family:Arial;color:rgb(0,0,0);font-style:italic;font-variant-nu=
meric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-s=
pace:pre-wrap">aggregators</span><span style=3D"font-size:10pt;font-family:=
Arial;color:rgb(0,0,0);font-variant-numeric:normal;font-variant-east-asian:=
normal;vertical-align:baseline;white-space:pre-wrap"> useful, but for those=
 that do implement this mitigation, sending the reports directly to all </s=
pan><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);font-s=
tyle:italic;font-variant-numeric:normal;font-variant-east-asian:normal;vert=
ical-align:baseline;white-space:pre-wrap">aggregators</span><span style=3D"=
font-size:10pt;font-family:Arial;color:rgb(0,0,0);font-variant-numeric:norm=
al;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-w=
rap"> presents some meaningful challenges.</span><br>Interesting, could you=
 be more specific? I&#39;m not aware of a DP mechanism for which option #2 =
would be a barrier.</blockquote><div><br></div><div>I&#39;m not suggesting =
that option #2 would prevent DP, but rather that a solution which employs a=
 DP mechanism to prevent the Sybil attack incurs the added operational cost=
s of option #2, without the added benefit.=C2=A0</div><div><br></div><div>C=
heers,</div><div>Erik</div></div><br><div class=3D"gmail_quote"><div dir=3D=
"ltr" class=3D"gmail_attr">On Wed, Jan 19, 2022 at 1:25 PM Christopher Patt=
on &lt;<a href=3D"mailto:cpatton@cloudflare.com">cpatton@cloudflare.com</a>=
&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px =
0px 0px 0.8ex;border-left-width:1px;border-left-style:solid;border-left-col=
or:rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div class=3D"gmail_=
quote"><div>Hi Erik!<br></div><div>=C2=A0</div><blockquote class=3D"gmail_q=
uote" style=3D"margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-s=
tyle:solid;border-left-color:rgb(204,204,204);padding-left:1ex"><div dir=3D=
"ltr"><span id=3D"gmail-m_-6392060375465505066m_1487501867054257734gmail-m_=
2460893677559864241gmail-docs-internal-guid-7091e80b-7fff-7c0e-b2b6-16a4405=
21498"><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;margin-botto=
m:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);bac=
kground-color:transparent;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">There is a 4th opti=
on being discussed in </span><a href=3D"https://github.com/abetterinternet/=
ppm-specification/issues/166" style=3D"text-decoration-line:none" target=3D=
"_blank"><span style=3D"font-size:10pt;font-family:Arial;background-color:t=
ransparent;font-variant-numeric:normal;font-variant-east-asian:normal;text-=
decoration-line:underline;vertical-align:baseline;white-space:pre-wrap">iss=
ue #166</span></a><span style=3D"font-size:10pt;font-family:Arial;color:rgb=
(0,0,0);background-color:transparent;font-variant-numeric:normal;font-varia=
nt-east-asian:normal;vertical-align:baseline;white-space:pre-wrap"> [1], wh=
ich moves in the opposite direction by proposing sending reports to a new p=
arty called a </span><span style=3D"font-size:10pt;font-family:Arial;color:=
rgb(0,0,0);background-color:transparent;font-style:italic;font-variant-nume=
ric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-spa=
ce:pre-wrap">consumer</span><span style=3D"font-size:10pt;font-family:Arial=
;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;=
font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap=
"> which is entirely untrusted and specified by the </span><span style=3D"f=
ont-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transpare=
nt;font-style:italic;font-variant-numeric:normal;font-variant-east-asian:no=
rmal;vertical-align:baseline;white-space:pre-wrap">collector</span><span st=
yle=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:t=
ransparent;font-variant-numeric:normal;font-variant-east-asian:normal;verti=
cal-align:baseline;white-space:pre-wrap">. Note that the </span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-style:italic;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">consumer</span><spa=
n style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-col=
or:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;v=
ertical-align:baseline;white-space:pre-wrap"> would also also be able to mo=
unt the Sybil attack, which I=E2=80=99ll address below.</span></p></span></=
div></blockquote><div><br></div><div>Let&#39;s call this option #5 for the =
sake of discussion. Option #4 is defined at the top of this thread.<br></di=
v>=C2=A0<br><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left-width:1px;border-left-style:solid;border-left-color:rgb(20=
4,204,204);padding-left:1ex"><div><span id=3D"gmail-m_-6392060375465505066m=
_1487501867054257734gmail-m_2460893677559864241gmail-docs-internal-guid-709=
1e80b-7fff-7c0e-b2b6-16a440521498"><p dir=3D"ltr" style=3D"line-height:1.38=
;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:10pt;font-famil=
y:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:=
normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:p=
re-wrap">The rationale here is almost entirely operational. Both of the API=
s above are implemented within an user-agent, but are used by arbitrary web=
sites and apps. Consuming and storing these reports sent from instances of =
those user-agents is a non-trivial infrastructure challenge, and we should =
assume that </span><span style=3D"font-size:10pt;font-family:Arial;color:rg=
b(0,0,0);background-color:transparent;font-style:italic;font-variant-numeri=
c:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space=
:pre-wrap">aggregators</span><span style=3D"font-size:10pt;font-family:Aria=
l;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal=
;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wra=
p"> will not have 100% uptime. In fact, improving from 99.9% uptime to 99.9=
9% uptime can often mean an order of magnitude increase in operational cost=
s.</span></p><br><p dir=3D"ltr" style=3D"line-height:1.38;margin-top:0pt;ma=
rgin-bottom:0pt"><span style=3D"font-size:10pt;font-family:Arial;color:rgb(=
0,0,0);background-color:transparent;font-variant-numeric:normal;font-varian=
t-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">I would a=
ssume that the set of parties offering their services as </span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-style:italic;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">aggregators</span><=
span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-=
color:transparent;font-variant-numeric:normal;font-variant-east-asian:norma=
l;vertical-align:baseline;white-space:pre-wrap"> will be fairly constrained=
 for the near-to-mid term, and as such, </span><span style=3D"font-size:10p=
t;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-styl=
e:italic;font-variant-numeric:normal;font-variant-east-asian:normal;vertica=
l-align:baseline;white-space:pre-wrap">collectors</span><span style=3D"font=
-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;=
font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:b=
aseline;white-space:pre-wrap"> will have minimal choice when it comes to fi=
nding the appropriate trade offs between operational concerns like uptime v=
s infrastructure costs. As such, I believe it=E2=80=99s beneficial to move =
as much of this cost and operational effort outside of the </span><span sty=
le=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tr=
ansparent;font-style:italic;font-variant-numeric:normal;font-variant-east-a=
sian:normal;vertical-align:baseline;white-space:pre-wrap">aggregator</span>=
</p></span></div></blockquote><div><br></div><div>The ENPA system called th=
is the &quot;ingestor&quot;. There the rationale was the same: To relieve t=
he aggregators of the operational burden of horizontally scaling to consume=
 client traffic. In the current spec, there is a distinguished aggregator, =
called the leader, who bears this responsibility.</div><div><br></div>I thi=
nk it makes sense for a leader deployment to have a separate &quot;leader-i=
ngestor&quot; service and &quot;leader-aggregator&quot; service that have d=
ifferent uptime requirements (and are perhaps maintained by different teams=
 within the same organization). My (current) objection to option #5 is that=
 roles like &quot;leader&quot;, &quot;helper&quot;, and &quot;collector&quo=
t; have more to do with the trust model than who is operationally responsib=
le for what. For instance, a statement &quot;We assume the leader and helpe=
r don&#39;t collude&quot; is meaningful because they get different sets of =
input shares; on the other hand, it doesn&#39;t sense to ask whether the le=
ader-ingestor and leader-aggregator collude, since they only see one set of=
 input shares.</div><div class=3D"gmail_quote"><br></div><div class=3D"gmai=
l_quote">It&#39;s just my opinion, but I think option #5 would only make se=
nse if the ingestor/consumer did something that is not already implemented =
by some server in the protocol.<br></div><div class=3D"gmail_quote"><div>=
=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left-width:1px;border-left-style:solid;border-left-color:rgb(20=
4,204,204);padding-left:1ex"><div dir=3D"ltr"><span id=3D"gmail-m_-63920603=
75465505066m_1487501867054257734gmail-m_2460893677559864241gmail-docs-inter=
nal-guid-7091e80b-7fff-7c0e-b2b6-16a440521498"><p dir=3D"ltr" style=3D"line=
-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=3D"font-size:10p=
t;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-vari=
ant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;w=
hite-space:pre-wrap">To the point on Sybil attacks: nearly all the APIs tha=
t would implement PPM will run in user-agents on untrusted and unknown devi=
ces. The code used to generate those reports will be open source, or at lea=
st the protocol will be publicly known, so the bigger threat, in my opinion=
, is a malicious entity operating a modified user-agent to generate events =
which enable a Sybil attack. The IPA proposal linked above (as well as the =
</span><a href=3D"https://github.com/WICG/conversion-measurement-api/blob/m=
ain/AGGREGATE.md" style=3D"text-decoration-line:none" target=3D"_blank"><sp=
an style=3D"font-size:10pt;font-family:Arial;background-color:transparent;f=
ont-variant-numeric:normal;font-variant-east-asian:normal;text-decoration-l=
ine:underline;vertical-align:baseline;white-space:pre-wrap">Attribution Rep=
orting API with Aggregate Reports</span></a><span style=3D"font-size:10pt;f=
ont-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant=
-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;whit=
e-space:pre-wrap"> [4]) propose using differential privacy to address these=
 types of attacks. This approach should also mitigate Sybil attacks.</span>=
</p></span></div></blockquote><div><br></div><div>DP has been discussed as =
a defense for Sybil attacks, as has (some variant of) Privacy Pass, which w=
ould allow the aggregators to avoid ingesting reports from clients they hav=
en&#39;t seen before. All of these things are still on the table, but so fa=
r have been considered as extensions to the core PPM protocol.<br></div><di=
v><br></div><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"mar=
gin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-style:solid;border-=
left-color:rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><span id=3D"=
gmail-m_-6392060375465505066m_1487501867054257734gmail-m_246089367755986424=
1gmail-docs-internal-guid-7091e80b-7fff-7c0e-b2b6-16a440521498"><p dir=3D"l=
tr" style=3D"line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical=
-align:baseline;white-space:pre-wrap">I don=E2=80=99t believe that all impl=
ementations will end up using differential privacy, and thus may find the p=
roposal for reports to be sent directly to all </span><span style=3D"font-s=
ize:10pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;fo=
nt-style:italic;font-variant-numeric:normal;font-variant-east-asian:normal;=
vertical-align:baseline;white-space:pre-wrap">aggregators</span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical=
-align:baseline;white-space:pre-wrap"> useful, but for those that do implem=
ent this mitigation, sending the reports directly to all </span><span style=
=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-color:tran=
sparent;font-style:italic;font-variant-numeric:normal;font-variant-east-asi=
an:normal;vertical-align:baseline;white-space:pre-wrap">aggregators</span><=
span style=3D"font-size:10pt;font-family:Arial;color:rgb(0,0,0);background-=
color:transparent;font-variant-numeric:normal;font-variant-east-asian:norma=
l;vertical-align:baseline;white-space:pre-wrap"> presents some meaningful c=
hallenges.</span></p></span></div></blockquote><div><br></div><div>Interest=
ing, could you be more specific? I&#39;m not aware of a DP mechanism for wh=
ich option #2 would be a barrier.<br></div></div></div>
</blockquote></div>

--000000000000fe2f0205d5f6f123--


From nobody Wed Jan 19 15:37:02 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 440F73A08D0 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 15:37:01 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id cuJa5R26x_Qb for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 15:36:58 -0800 (PST)
Received: from mail-pj1-x102f.google.com (mail-pj1-x102f.google.com [IPv6:2607:f8b0:4864:20::102f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 4E0283A08D2 for <ppm@ietf.org>; Wed, 19 Jan 2022 15:36:58 -0800 (PST)
Received: by mail-pj1-x102f.google.com with SMTP id g11-20020a17090a7d0b00b001b2c12c7273so3511341pjl.0 for <ppm@ietf.org>; Wed, 19 Jan 2022 15:36:58 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=qQcdTOswRh+bpyDc/dE7384h5i1u96fOMgjGOESUtHA=; b=UErI66PWkXRjOe59lhSHA0Y8OgID5L4ijPQr4MYKAPKFyTkjFucxtdGSJMUpTj/63d fINAAm/Ri1JCRqvnthLbCdAoyE0de1aCgD/q9xFfgalNlWDHXSdDhv5skLzy4NLIugl7 gJ1TfYoqxeO39QhMC+gRS6/4iMFTtkR1CUVtg=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=qQcdTOswRh+bpyDc/dE7384h5i1u96fOMgjGOESUtHA=; b=nt00kRINVOoox5ynOzshnsqEnGWa1782dK9l0uDYMuKFqNtisfOdE4zcr4PAb7Nm3S 39zTesVh6KJ6+Kun2UMq1MtuAG2V67vu0KzSc4C8mybm2a17VBF20rai6PbMbkFumCcq r2Wm3caa2hvybQPYkGi/fP6NJUJFdQ+DtTNcZZ+ZG//clKm8DpbqYW0t3wV//b/Mig4o fieqrPZUMNX9SGC4cxBjPWzkUl4jp6glL2n+rttkqCuymhHV9G5zZ1c8lr7dZA3uEQmt Rfn+IDbPBWCy2f/JzbnGy+tKOYlkVx8S1riNV0RMRN/B4SxXc0J134/xHDdhXAbQ5eVT cCQg==
X-Gm-Message-State: AOAM532QiSSM69hQqflKG3VQ9QRuoaXHFSsZX7ghwBCrl/XHnA58vOmH v9UnsW/cr4ECxr4KyQYat7WkcnX1+C/rQ+n+nQLWDQ==
X-Google-Smtp-Source: ABdhPJxTfSswc0lSDpFnsL5rQLK8GhnsdIkCqRYw26YN78fk1hCAkf8DV7bB13cklALZ2JcqBrYqvYiCzlhoBt9Hpcs=
X-Received: by 2002:a17:90a:9f81:: with SMTP id o1mr7192750pjp.240.1642635415728;  Wed, 19 Jan 2022 15:36:55 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com>
In-Reply-To: <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Wed, 19 Jan 2022 15:36:44 -0800
Message-ID: <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Charlie Harrison <csharrison@google.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000073a32705d5f7dcbb"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/lZMVdcU3ONWK2WfAVhhFH-pUtHQ>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 19 Jan 2022 23:37:01 -0000

--00000000000073a32705d5f7dcbb
Content-Type: text/plain; charset="UTF-8"

Erik,


> What I am proposing, and is discussed in issue #166 and #64, is allowing
> an entirely untrusted entity to play the role of "ingestor", so that a
> "collector" can either run this system themselves, or be able to hire any
> service (beyond the set of approved organizations that run aggregators.)
> This is particularly important in the setting where a web platform vendor
> may have a very small set of approved organizations; shifting as much of
> the operational cost out of that small set.
>
> From the "collector" point of view, I see these options as
> Option #1: I have a limited set of leaders to choose from which can
> receive my reports, without much optionality between uptime and cost.
> Option #2: I have an even more limited set of aggregators which must
> receive my reports, and uptime is dependent on all of them.
> Option #5: I can choose to ingest reports myself, or outsource it to any
> service provider. I still have a limited set of aggregators, but I only
> rely on them for aggregation and not real-time ingestion.
>


> If we can design solutions in which option #5 still provides the intended
> security and privacy properties, it seems greatly preferable to the
> "collector". That said, I am advocating for option #3, with the choice that
> is inclusive of option #5.
>

This use case makes perfect sense to me, and I think we should make sure
the protocol is written in a way that enables it. I'm just not sure doing
so necessitates option #5, i.e., defining an explicit ingestor role. In
particular, if we take option #1, there is nothing that stops the
collector/ingestor from uploading the reports to the leader directly. The
only difference is that the leader may wish to authenticate the upload
request (as is done in ENPA for instance), but this is a matter for the
deployment and needn't be addressed in the core PPM protocol. (Of course,
if we need to define an extension to support this use case, then we can and
should do so.)

The same seems true if we take option #2. Here I would just say that
standing up an ingestion server between the clients and the aggregators is
even more desirable, since the helper wouldn't have to absorb client
traffic. But here again I view this as a matter for the deployment and not
something that impacts the core protocol logic.

Perhaps one way to resolve this question is to have the document enumerate
the various deployment scenarios we envision and make sure that the core
protocol logic supports all of them.


There is a meaningful set of attacks where the "ingestor" is or colludes
> with the "collector", given that the collector may have added context. For
> example, in this case the "collector" and "ingestor" could work together to
> mount the Sybil attack you describe.
>

Ah, true enough! Though we've been considering an even stronger attacker
for privacy: One that controls the collector completely and controls a
subset of the aggregators (but not all of them). This is somewhere in
Security Considerations, but I don't think this is spelled out that well.



> I'm not suggesting that option #2 would prevent DP, but rather that a
> solution which employs a DP mechanism to prevent the Sybil attack incurs
> the added operational costs of option #2, without the added benefit.
>

Can you provide more detail? What benefit is lost, exactly? Do you have an
attack? I want to drill down on this because it sounds like a potential
deal breaker for #2. (More broadly, any option that requires us to
significantly compromise security should be off the table.)


Chris P.

--00000000000073a32705d5f7dcbb
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div class=3D"gmail_quote">Erik,<br></div><div class=3D"gm=
ail_quote"><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"marg=
in:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1e=
x"><div dir=3D"ltr"><div></div><div>What I am proposing, and is discussed=
=C2=A0in issue #166 and #64, is allowing an entirely untrusted entity to pl=
ay the role of &quot;ingestor&quot;, so that a &quot;collector&quot; can ei=
ther run this system themselves, or be able to hire any service (beyond the=
 set of approved organizations that run aggregators.) This is particularly =
important in the setting where a web platform vendor may have a very small =
set of approved organizations; shifting as much of the operational cost out=
 of that small set.</div><div><br></div><div>From the &quot;collector&quot;=
 point of view, I see these options as</div><div>Option #1: I have a limite=
d set of leaders to choose from which can receive my reports, without much =
optionality between uptime and cost.</div><div>Option #2: I have an even mo=
re limited set of aggregators which must receive my reports, and uptime is =
dependent on all of them.</div><div>Option #5: I can choose to ingest repor=
ts myself, or outsource it to any service provider. I still have a limited =
set of aggregators, but I only rely on them for aggregation and not real-ti=
me ingestion.</div></div></blockquote><div><div>=C2=A0</div><blockquote cla=
ss=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid =
rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>If
 we can design solutions in which option #5 still provides the intended=20
security and privacy properties, it seems greatly preferable to the=20
&quot;collector&quot;. That said, I am advocating for option #3, with the c=
hoice=20
that is inclusive of option #5.</div></div></blockquote></div><div><br></di=
v><div>This use case makes perfect sense to me, and I think we should make =
sure the protocol is written in a way that enables it. I&#39;m just not sur=
e doing so necessitates option #5, i.e., defining an explicit ingestor role=
. In particular, if we take option #1, there is nothing that stops the coll=
ector/ingestor from uploading the reports to the leader directly. The only =
difference is that the leader may wish to authenticate the upload request (=
as is done in ENPA for instance), but this is a matter for the deployment a=
nd needn&#39;t be addressed in the core PPM protocol. (Of course, if we nee=
d to define an extension to support this use case, then we can and should d=
o so.)</div><div><br></div><div>The same seems true if we take option #2. H=
ere I would just say that standing up an ingestion server between the clien=
ts and the aggregators is even more desirable, since the helper wouldn&#39;=
t have to absorb client traffic. But here again I view this as a matter for=
 the deployment and not something that impacts the core protocol logic.</di=
v><div><br></div><div>Perhaps one way to resolve this question is to have t=
he document enumerate the various deployment scenarios we envision and make=
 sure that the core protocol logic supports all of them. <br></div><div><br=
></div><div><br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px=
 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><di=
v dir=3D"ltr"><div>There is a meaningful set of attacks where the &quot;ing=
estor&quot; is or colludes with the &quot;collector&quot;, given that the c=
ollector may have added context. For example, in this case the &quot;collec=
tor&quot; and &quot;ingestor&quot; could work together to mount the Sybil a=
ttack you describe.</div></div></blockquote><div><br></div><div>Ah, true en=
ough! Though we&#39;ve been considering an even stronger attacker for priva=
cy: One that controls the collector completely and controls a subset of the=
 aggregators (but not all of them). This is somewhere in Security Considera=
tions, but I don&#39;t think this is spelled out that well.</div><div><br><=
/div><div></div><div>=C2=A0 <br></div><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div dir=3D"ltr"><div></div><div>I&#39;m not suggesting that o=
ption #2 would prevent DP, but rather that a solution which employs a DP me=
chanism to prevent the Sybil attack incurs the added operational costs of o=
ption #2, without the added benefit.=C2=A0</div></div></blockquote><div><br=
></div><div>Can you provide more detail? What benefit is lost, exactly? Do =
you have an attack? I want to drill down on this because it sounds like a p=
otential deal breaker for #2. (More broadly, any option that requires us to=
 significantly compromise security should be off the table.)</div><div><br>=
</div><div><br></div><div>Chris P.<br></div></div></div>

--00000000000073a32705d5f7dcbb--


From nobody Wed Jan 19 16:52:12 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 58A263A109A for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 16:52:11 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id XXEdvHR8OgW6 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 16:52:07 -0800 (PST)
Received: from mail-lf1-x12e.google.com (mail-lf1-x12e.google.com [IPv6:2a00:1450:4864:20::12e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id A93413A1094 for <ppm@ietf.org>; Wed, 19 Jan 2022 16:52:06 -0800 (PST)
Received: by mail-lf1-x12e.google.com with SMTP id h26so474762lfv.11 for <ppm@ietf.org>; Wed, 19 Jan 2022 16:52:06 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=t66p+FwLu7jZ1aKkZKGBp6rodO05hnCH/PRKJkyI7AY=; b=f5/S68Eb1xTfa4RvHxbwIrhwrgCb7EWQGmzeQjTn6LwecGBD3jWIXvM447M7nGZIoV f4Q05WvYOQfsi7bpuNGhke3Y77+t4pLUaR660mBqMZ6pukhxaEilnDtayNYpRJCwO+YO 65pZlEazXl1cvAVAUv1FfHLL45B+tF1ZB+3L/rv+kTerl52iTEESW1330rQZNqtbv5io cpwq+tjrDAKGRkJh5VCuCamPHpMKzX57MDcUWc2RpQGWALn+pgokSIJeyAuRCLU5lAsk q8qP2IjIMa5e3lsndduE18B8gdTZAyZLN3BPNrJWt3+x0gaxl5kiglL1UcVxh/ZMVR53 Dyew==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=t66p+FwLu7jZ1aKkZKGBp6rodO05hnCH/PRKJkyI7AY=; b=mVW/T06l9JQ4OfYgE5VGYJQTqudTdx6fzt1z87yolMGTe21ntK7Yo9DVDvqIZQ/ojp Vkd9TebCjfRgrO3cZi8vWSTCarFKUZziEEKdI1m8bM+/HquJ7czqYVMyKJ5z8c+KVzI6 SErujbxIYEG2TDG3cBNSgUnjzLrZNagapXDFR0g2N8mLgOoczo6qiIWQFByy9CGk/tbP 5i+JISEk5p41zohza2EhoBCFVzYPLBtbLTHVGf4L9XkolsMKfyZVbmWUYqsUaQtmFApH r9YHpjFcNXVl/+zZNbB3XUsmABPb1L33mNYUveMSd90+tBiDMTznkjMoz2dN/IeBYn6/ JJEQ==
X-Gm-Message-State: AOAM532oFjKsbH1qOBOuuIfzDeVyI9GQfUHo+SPnJfLg4syGlAr5E0QC b9nJ9vujHO7BXyWLUvOHlqwZMFoJ0BBbsLNPIOY=
X-Google-Smtp-Source: ABdhPJzHc2f/Zl/rV+H0PcxSRb1al2PbiW+lsF1zVSdvxPXTJUFBUdjdpdeaju22EkBpX/Cc2fvVcmvObbk6lpsO5t0=
X-Received: by 2002:a2e:9c88:: with SMTP id x8mr24212793lji.111.1642639923443;  Wed, 19 Jan 2022 16:52:03 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com>
In-Reply-To: <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Wed, 19 Jan 2022 16:51:52 -0800
Message-ID: <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com>
To: Christopher Patton <cpatton@cloudflare.com>
Cc: Charlie Harrison <csharrison@google.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000217d8605d5f8e912"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/4ju_voQRIvQ0x2r3G6XOcoqCVtE>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 00:52:12 -0000

--000000000000217d8605d5f8e912
Content-Type: text/plain; charset="UTF-8"

Chris,
 Thanks for your patient feedback, I might be catching up a bit here, but
it's been very helpful for me to wrap my head around this.

This use case makes perfect sense to me, and I think we should make sure
> the protocol is written in a way that enables it.


Great! My initial read was that it wouldn't currently enable it, and I'm
happy to work on an issue/PR that can make that more clear and enumerate
various deployment scenarios, as you suggest. One point of confusion was
the suggestion about limiting reports from a single IP address, as I'd
expect an intermediary server between the client and leader/aggregators to
likely have a fixed IP address.

The same seems true if we take option #2. Here I would just say that
> standing up an ingestion server between the clients and the aggregators is
> even more desirable, since the helper wouldn't have to absorb client
> traffic.


Agreed. My misunderstanding was that this would require all aggregators to
handle real-time client traffic. If this can happen as a batch from an
ingestion service, having it go directly to each aggregator seems
preferable.

Can you provide more detail? What benefit is lost, exactly? Do you have an
> attack? I want to drill down on this because it sounds like a potential
> deal breaker for #2. (More broadly, any option that requires us to
> significantly compromise security should be off the table.)


I was not envisioning an attack, but rather that all aggregators would need
to handle real-time client traffic, which could create quite a bit of
operational risk. But given that was my misunderstanding, I don't think
this is a detractor for option #2.

Cheers,
Erik



On Wed, Jan 19, 2022 at 3:36 PM Christopher Patton <cpatton@cloudflare.com>
wrote:

> Erik,
>
>
>> What I am proposing, and is discussed in issue #166 and #64, is allowing
>> an entirely untrusted entity to play the role of "ingestor", so that a
>> "collector" can either run this system themselves, or be able to hire any
>> service (beyond the set of approved organizations that run aggregators.)
>> This is particularly important in the setting where a web platform vendor
>> may have a very small set of approved organizations; shifting as much of
>> the operational cost out of that small set.
>>
>> From the "collector" point of view, I see these options as
>> Option #1: I have a limited set of leaders to choose from which can
>> receive my reports, without much optionality between uptime and cost.
>> Option #2: I have an even more limited set of aggregators which must
>> receive my reports, and uptime is dependent on all of them.
>> Option #5: I can choose to ingest reports myself, or outsource it to any
>> service provider. I still have a limited set of aggregators, but I only
>> rely on them for aggregation and not real-time ingestion.
>>
>
>
>> If we can design solutions in which option #5 still provides the intended
>> security and privacy properties, it seems greatly preferable to the
>> "collector". That said, I am advocating for option #3, with the choice that
>> is inclusive of option #5.
>>
>
> This use case makes perfect sense to me, and I think we should make sure
> the protocol is written in a way that enables it. I'm just not sure doing
> so necessitates option #5, i.e., defining an explicit ingestor role. In
> particular, if we take option #1, there is nothing that stops the
> collector/ingestor from uploading the reports to the leader directly. The
> only difference is that the leader may wish to authenticate the upload
> request (as is done in ENPA for instance), but this is a matter for the
> deployment and needn't be addressed in the core PPM protocol. (Of course,
> if we need to define an extension to support this use case, then we can and
> should do so.)
>
> The same seems true if we take option #2. Here I would just say that
> standing up an ingestion server between the clients and the aggregators is
> even more desirable, since the helper wouldn't have to absorb client
> traffic. But here again I view this as a matter for the deployment and not
> something that impacts the core protocol logic.
>
> Perhaps one way to resolve this question is to have the document enumerate
> the various deployment scenarios we envision and make sure that the core
> protocol logic supports all of them.
>
>
> There is a meaningful set of attacks where the "ingestor" is or colludes
>> with the "collector", given that the collector may have added context. For
>> example, in this case the "collector" and "ingestor" could work together to
>> mount the Sybil attack you describe.
>>
>
> Ah, true enough! Though we've been considering an even stronger attacker
> for privacy: One that controls the collector completely and controls a
> subset of the aggregators (but not all of them). This is somewhere in
> Security Considerations, but I don't think this is spelled out that well.
>
>
>
>> I'm not suggesting that option #2 would prevent DP, but rather that a
>> solution which employs a DP mechanism to prevent the Sybil attack incurs
>> the added operational costs of option #2, without the added benefit.
>>
>
> Can you provide more detail? What benefit is lost, exactly? Do you have an
> attack? I want to drill down on this because it sounds like a potential
> deal breaker for #2. (More broadly, any option that requires us to
> significantly compromise security should be off the table.)
>
>
> Chris P.
>

--000000000000217d8605d5f8e912
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Chris,</div><div>=C2=A0Thanks for your patient feedba=
ck, I might be catching up a bit here, but it&#39;s been very helpful for m=
e to wrap my head around this.</div><div><br></div><blockquote class=3D"gma=
il_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,2=
04,204);padding-left:1ex">This use case makes perfect sense to me, and I th=
ink we should make sure the protocol is written in a way that enables it.=
=C2=A0</blockquote><div><br></div><div>Great! My initial read was that it w=
ouldn&#39;t currently enable it, and I&#39;m happy to work on an issue/PR t=
hat can make that more clear and enumerate various deployment=C2=A0scenario=
s, as you suggest. One point of confusion was the suggestion about limiting=
 reports from a single IP address, as I&#39;d expect an intermediary server=
 between the client and leader/aggregators to likely have a fixed IP addres=
s.</div><div><br></div><blockquote class=3D"gmail_quote" style=3D"margin:0p=
x 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Th=
e same seems true if we take option #2. Here I would just say that standing=
 up an ingestion server between the clients and the aggregators is even mor=
e desirable, since the helper wouldn&#39;t have to absorb client traffic.</=
blockquote><div><br></div><div>Agreed. My misunderstanding was that this wo=
uld require all aggregators to handle real-time client=C2=A0traffic. If thi=
s can happen as a batch from an ingestion service, having it go directly to=
 each aggregator seems preferable.</div><div><br></div><blockquote class=3D=
"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(2=
04,204,204);padding-left:1ex">Can you provide more detail? What benefit is =
lost, exactly? Do you have an attack? I want to drill down on this because =
it sounds like a potential deal breaker for #2. (More broadly, any option t=
hat requires us to significantly compromise security should be off the tabl=
e.)</blockquote><div><br></div><div>I was not envisioning an attack, but ra=
ther that all aggregators=C2=A0would need to handle real-time client traffi=
c, which could create quite a bit of operational risk. But given that was m=
y misunderstanding, I don&#39;t think this is a detractor for option #2.</d=
iv><div><br></div><div>Cheers,</div><div>Erik</div><div><br></div><div>=C2=
=A0</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Wed, Jan 19, 2022 at 3:36 PM Christopher Patton &lt;<a href=3D"=
mailto:cpatton@cloudflare.com">cpatton@cloudflare.com</a>&gt; wrote:<br></d=
iv><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bord=
er-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div =
class=3D"gmail_quote">Erik,<br></div><div class=3D"gmail_quote"><div>=C2=A0=
</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;b=
order-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><d=
iv></div><div>What I am proposing, and is discussed=C2=A0in issue #166 and =
#64, is allowing an entirely untrusted entity to play the role of &quot;ing=
estor&quot;, so that a &quot;collector&quot; can either run this system the=
mselves, or be able to hire any service (beyond the set of approved organiz=
ations that run aggregators.) This is particularly important in the setting=
 where a web platform vendor may have a very small set of approved organiza=
tions; shifting as much of the operational cost out of that small set.</div=
><div><br></div><div>From the &quot;collector&quot; point of view, I see th=
ese options as</div><div>Option #1: I have a limited set of leaders to choo=
se from which can receive my reports, without much optionality between upti=
me and cost.</div><div>Option #2: I have an even more limited set of aggreg=
ators which must receive my reports, and uptime is dependent on all of them=
.</div><div>Option #5: I can choose to ingest reports myself, or outsource =
it to any service provider. I still have a limited set of aggregators, but =
I only rely on them for aggregation and not real-time ingestion.</div></div=
></blockquote><div><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex"><div dir=3D"ltr"><div>If
 we can design solutions in which option #5 still provides the intended=20
security and privacy properties, it seems greatly preferable to the=20
&quot;collector&quot;. That said, I am advocating for option #3, with the c=
hoice=20
that is inclusive of option #5.</div></div></blockquote></div><div><br></di=
v><div>This use case makes perfect sense to me, and I think we should make =
sure the protocol is written in a way that enables it. I&#39;m just not sur=
e doing so necessitates option #5, i.e., defining an explicit ingestor role=
. In particular, if we take option #1, there is nothing that stops the coll=
ector/ingestor from uploading the reports to the leader directly. The only =
difference is that the leader may wish to authenticate the upload request (=
as is done in ENPA for instance), but this is a matter for the deployment a=
nd needn&#39;t be addressed in the core PPM protocol. (Of course, if we nee=
d to define an extension to support this use case, then we can and should d=
o so.)</div><div><br></div><div>The same seems true if we take option #2. H=
ere I would just say that standing up an ingestion server between the clien=
ts and the aggregators is even more desirable, since the helper wouldn&#39;=
t have to absorb client traffic. But here again I view this as a matter for=
 the deployment and not something that impacts the core protocol logic.</di=
v><div><br></div><div>Perhaps one way to resolve this question is to have t=
he document enumerate the various deployment scenarios we envision and make=
 sure that the core protocol logic supports all of them. <br></div><div><br=
></div><div><br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px=
 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><di=
v dir=3D"ltr"><div>There is a meaningful set of attacks where the &quot;ing=
estor&quot; is or colludes with the &quot;collector&quot;, given that the c=
ollector may have added context. For example, in this case the &quot;collec=
tor&quot; and &quot;ingestor&quot; could work together to mount the Sybil a=
ttack you describe.</div></div></blockquote><div><br></div><div>Ah, true en=
ough! Though we&#39;ve been considering an even stronger attacker for priva=
cy: One that controls the collector completely and controls a subset of the=
 aggregators (but not all of them). This is somewhere in Security Considera=
tions, but I don&#39;t think this is spelled out that well.</div><div><br><=
/div><div></div><div>=C2=A0 <br></div><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div dir=3D"ltr"><div></div><div>I&#39;m not suggesting that o=
ption #2 would prevent DP, but rather that a solution which employs a DP me=
chanism to prevent the Sybil attack incurs the added operational costs of o=
ption #2, without the added benefit.=C2=A0</div></div></blockquote><div><br=
></div><div>Can you provide more detail? What benefit is lost, exactly? Do =
you have an attack? I want to drill down on this because it sounds like a p=
otential deal breaker for #2. (More broadly, any option that requires us to=
 significantly compromise security should be off the table.)</div><div><br>=
</div><div><br></div><div>Chris P.<br></div></div></div>
</blockquote></div>

--000000000000217d8605d5f8e912--


From nobody Wed Jan 19 17:28:50 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id D92423A12B3 for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 17:28:48 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id e33AsRXA-gtC for <ppm@ietfa.amsl.com>; Wed, 19 Jan 2022 17:28:44 -0800 (PST)
Received: from mail-pj1-x1032.google.com (mail-pj1-x1032.google.com [IPv6:2607:f8b0:4864:20::1032]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id A6E2C3A12B0 for <ppm@ietf.org>; Wed, 19 Jan 2022 17:28:44 -0800 (PST)
Received: by mail-pj1-x1032.google.com with SMTP id s2-20020a17090ad48200b001b501977b23so3417851pju.2 for <ppm@ietf.org>; Wed, 19 Jan 2022 17:28:44 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=AbibboRlUMV0zlA2CCVVYGv8fj7G5YHPbGV/GIY0aHY=; b=xsxPQnLPpQFInX/Epr0jww0CW4MbjruahT0pSH+5To9F3YzQqZ2WY6qUuB4IfAMXx+ mvOx7Y6S61ukJijPkl7o1qwCbJ3Jzi2EH1XHsSAFBtlqMZl8DjwRdprEZYT+D4KjD2IQ cabEfkxIGST+jOQDuCjWNIJKK/D92Nz5EWsFc=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=AbibboRlUMV0zlA2CCVVYGv8fj7G5YHPbGV/GIY0aHY=; b=XR8AzHdW1Cvgbj9TjrqMa+CI+vNYSiQEAO+68VhHU7J7Bx3wyu7Op92dMoi2fK7QpP zKfghYNM1NXiTKXVz+jifoShHeHISqxAfjO6JT4qNEhAK5IDWQW744obQ7XFn0ku5/oy 6GDKEACQ70aGGlnT0ivwx4FHpGIkT8OPUvTH0S7uUbOwWjXEyxc5/4qsiea0IbqPopqa Hp97u4YCw1n74N+VcX3yucLEyEQs/NVD9dJImbHIHeSIrArfwL7O96uKkN8IyrXI15L9 XNVM71lDR6lgnJW2TFVTKiNlgM2N31Av8tV4a51ZwleQEQtFF/97GTvk03DWLYhCPQPA YXCw==
X-Gm-Message-State: AOAM5334rsmIKUdROjU2QQYWw6ii0qr1BJ4Aocj1hfdvtFUO4pVcmHpo d6IXsqMHlZIathvXEFSimHHgGZGTuCgCm5hX3KCpYA==
X-Google-Smtp-Source: ABdhPJxarXiTBU6fkiMG+l2+NetX25ReLYkT2U3fKTbEJo0+7tyuvkOfa0YGRi6e9PEQMPsTkaQcJGhxQ+ZvU6mEqs0=
X-Received: by 2002:a17:90a:9f81:: with SMTP id o1mr7695240pjp.240.1642642122806;  Wed, 19 Jan 2022 17:28:42 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com>
In-Reply-To: <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Wed, 19 Jan 2022 17:28:31 -0800
Message-ID: <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Charlie Harrison <csharrison@google.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000393da605d5f96c94"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/rFCj4pRPp-c2Rp_qljBzwLSLQzI>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 01:28:49 -0000

--000000000000393da605d5f96c94
Content-Type: text/plain; charset="UTF-8"

> Great! My initial read was that it wouldn't currently enable it, and I'm
> happy to work on an issue/PR that can make that more clear and enumerate
> various deployment scenarios, as you suggest. One point of confusion was
> the suggestion about limiting reports from a single IP address, as I'd
> expect an intermediary server between the client and leader/aggregators to
> likely have a fixed IP address.
>

Awesome! Let's start with opening an issue and folks to start putting their
deployment setups there.

Yes, rate-limiting clients based on IP alone is impossible for aggregators
if they're behind a proxy. It should be noted that this is a heuristic only
that an attacker may be able to defeat in many ways. To deal with with
Sybil attacks in general, we'll definitely have to do a bit more anyway.


Chris P.

--000000000000393da605d5f96c94
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><br>=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"=
gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(20=
4,204,204);padding-left:1ex"><div dir=3D"ltr"><div></div><div>Great! My ini=
tial read was that it wouldn&#39;t currently enable it, and I&#39;m happy t=
o work on an issue/PR that can make that more clear and enumerate various d=
eployment=C2=A0scenarios, as you suggest. One point of confusion was the su=
ggestion about limiting reports from a single IP address, as I&#39;d expect=
 an intermediary server between the client and leader/aggregators to likely=
 have a fixed IP address.</div></div></blockquote><div><br></div><div>Aweso=
me! Let&#39;s start with opening an issue and folks to start putting their =
deployment setups there.</div><div><br></div><div>Yes, rate-limiting client=
s based on IP alone is impossible for aggregators if they&#39;re behind a p=
roxy. It should be noted that this is a heuristic only that an attacker may=
 be able to defeat in many ways. To deal with with Sybil attacks in general=
, we&#39;ll definitely have to do a bit more anyway.<br></div>=C2=A0</div><=
div class=3D"gmail_quote"><br></div><div class=3D"gmail_quote">Chris P.<br>=
</div></div>

--000000000000393da605d5f96c94--


From nobody Thu Jan 20 08:04:09 2022
Return-Path: <csharrison@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 350C43A1727 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 08:04:07 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id m8Mf8pguNNFM for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 08:04:03 -0800 (PST)
Received: from mail-ed1-x52a.google.com (mail-ed1-x52a.google.com [IPv6:2a00:1450:4864:20::52a]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id C12A43A1726 for <ppm@ietf.org>; Thu, 20 Jan 2022 08:04:02 -0800 (PST)
Received: by mail-ed1-x52a.google.com with SMTP id c24so28381584edy.4 for <ppm@ietf.org>; Thu, 20 Jan 2022 08:04:02 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=ZZugEprKUoRLJKXKH0S2pS81GoxgjfIux2VsvI8w5Zg=; b=P3JivNCKTWb9FvtH1NhS2qa3y/lP+43ifWz7GefuT+IvO8dUL2w9r0+l//EYXxBHYY gHah95kzVCS1CYuJkusBlbemGx6iLeH9KAXUe70Fhl4CCQjPEShhG822MB5PuL267DVo oJOYONE7z37WMuWfJ0rLrRBZbaVSiNJm5/fyGiY5xNiih1TK37riS0tLPsCXjyDzb0bz imLnz1px7YvJYkh1t3FOFLpAohnK3jK9Tb+NAj3TE4Mz4Fq2DchJUHdiFKPukt1CZG1u 4ZE8dftsRNKAsVAm2TM162/kOcxjeGP4KHw+6l+Ef9GTAVypHBtIswkpNZoNraGSOfqM qlzw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=ZZugEprKUoRLJKXKH0S2pS81GoxgjfIux2VsvI8w5Zg=; b=vVc0Us1Qo4lTXJMxGSx6MiG7NXV0GMBnEobNCVMgHdz7Qg2uaIDyX+SYIIk50vNe4/ B9/Nw/uvXUQdKpmGDas/0SkyL0YM9QKxNSLkYNo/1ZK84xds6RweRWFf9iT+2D543WRv bz/OjPR72O1ZCK1mkeLTUjrBrUUYNrZGjDcBH80AhGo2Lrju20YrPtV113OiphoPk5+5 NkCKIC/d6VvXYYZWbO+q1fKymMhSEiChodboLyi09J6uA3Aw+HwmrTRMjkbkP6pUffh1 w7dZ+xiMeyKwFXfh4gpeEpXs72eHSqUPd5c29+hf8HJVrC5r+D8jVgIXEjyW92rN4Evd B6LA==
X-Gm-Message-State: AOAM533Hfajh2DuuDhzk56a3WGTwgXdiY8yZt7IpyGueU8xcyl2Lg0+5 bbysZUvoLD42W7XTvKaC5lPMlOozdWmTGMK5D2pIqw==
X-Google-Smtp-Source: ABdhPJyvSLHVEQWjIRjBMTEt2mNtHcc0P2/Wd9kyE3zuIA27Q0AMGTIQfr5cpptXe9Rzj5qUS/kTqRmY8VMZ3rjpP9Q=
X-Received: by 2002:a05:6402:11c9:: with SMTP id j9mr37188871edw.411.1642694639470;  Thu, 20 Jan 2022 08:03:59 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com>
In-Reply-To: <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com>
From: Charlie Harrison <csharrison@google.com>
Date: Thu, 20 Jan 2022 10:03:48 -0600
Message-ID: <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com>
To: Christopher Patton <cpatton@cloudflare.com>
Cc: Erik Taubeneck <erik.taubeneck@gmail.com>, Phillipp Schoppmann <schoppmann@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000007652be05d605a61c"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/4QqSdOsALDC4fLqp0DLoXxoNp-c>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 16:04:07 -0000

--0000000000007652be05d605a61c
Content-Type: text/plain; charset="UTF-8"

Thanks Chris,
I think I am in agreement with Erik that we could mitigate the issues with
Option #2 if there is an ingestor client. I think framing the protocol this
way actually allows close imitation of option #1 and #2

- An "option 1" deployment could just specify one of the aggregators to
work as the ingestor even if it isn't required by the protocol. They just
happen to be the same party.
- An "option 2" deployment could have the ingestor fan out reports to
multiple aggregators (or ignore the ingestor completely)

Your other points make perfect sense to me!

On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <cpatton@cloudflare.com>
wrote:

>
>
>
>> Great! My initial read was that it wouldn't currently enable it, and I'm
>> happy to work on an issue/PR that can make that more clear and enumerate
>> various deployment scenarios, as you suggest. One point of confusion was
>> the suggestion about limiting reports from a single IP address, as I'd
>> expect an intermediary server between the client and leader/aggregators to
>> likely have a fixed IP address.
>>
>
> Awesome! Let's start with opening an issue and folks to start putting
> their deployment setups there.
>
> Yes, rate-limiting clients based on IP alone is impossible for aggregators
> if they're behind a proxy. It should be noted that this is a heuristic only
> that an attacker may be able to defeat in many ways. To deal with with
> Sybil attacks in general, we'll definitely have to do a bit more anyway.
>
>
> Chris P.
>

--0000000000007652be05d605a61c
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Thanks Chris,<div>I think I am in agreement with Erik that=
 we could mitigate the issues with Option #2 if there is an ingestor client=
. I think framing the protocol this way actually allows close imitation=C2=
=A0of option #1 and #2</div><div><br></div><div>- An &quot;option 1&quot; d=
eployment could=C2=A0just=C2=A0specify one of the aggregators to work as th=
e ingestor even if it isn&#39;t required by the protocol. They just happen =
to be the same party.</div><div>- An &quot;option 2&quot; deployment could =
have the ingestor fan out reports to multiple aggregators (or ignore the in=
gestor completely)</div><div><br></div><div>Your other points make perfect =
sense to me!</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" cla=
ss=3D"gmail_attr">On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a=
 href=3D"mailto:cpatton@cloudflare.com">cpatton@cloudflare.com</a>&gt; wrot=
e:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"l=
tr"><br>=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" =
style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);pa=
dding-left:1ex"><div dir=3D"ltr"><div></div><div>Great! My initial read was=
 that it wouldn&#39;t currently enable it, and I&#39;m happy to work on an =
issue/PR that can make that more clear and enumerate various deployment=C2=
=A0scenarios, as you suggest. One point of confusion was the suggestion abo=
ut limiting reports from a single IP address, as I&#39;d expect an intermed=
iary server between the client and leader/aggregators to likely have a fixe=
d IP address.</div></div></blockquote><div><br></div><div>Awesome! Let&#39;=
s start with opening an issue and folks to start putting their deployment s=
etups there.</div><div><br></div><div>Yes, rate-limiting clients based on I=
P alone is impossible for aggregators if they&#39;re behind a proxy. It sho=
uld be noted that this is a heuristic only that an attacker may be able to =
defeat in many ways. To deal with with Sybil attacks in general, we&#39;ll =
definitely have to do a bit more anyway.<br></div>=C2=A0</div><div class=3D=
"gmail_quote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>

--0000000000007652be05d605a61c--


From nobody Thu Jan 20 09:20:03 2022
Return-Path: <schoppmann@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id EC75B3A19E4 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 09:19:53 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -22.599
X-Spam-Level: 
X-Spam-Status: No, score=-22.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id pTsA5IigQvtF for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 09:19:49 -0800 (PST)
Received: from mail-wm1-x32f.google.com (mail-wm1-x32f.google.com [IPv6:2a00:1450:4864:20::32f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id EF96A3A146B for <ppm@ietf.org>; Thu, 20 Jan 2022 09:19:40 -0800 (PST)
Received: by mail-wm1-x32f.google.com with SMTP id w26so13700063wmi.0 for <ppm@ietf.org>; Thu, 20 Jan 2022 09:19:40 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=L2rribcOjnwt4VNORgwAor4H7h6UauEEJ+gLIsmop6Y=; b=sLZTid1VtysakCzEcT9zF5ZzhIOgirfxIWfFz9wDZvqetGJ+63gk48CYwW7R6k5uQm H73/r4lTjWUGCZcexfJEeEzqRbzj1HqPt3s6ol7eQWJ/LmAf51ytRrDU33hQ9Ywf9dwe lhBaQGGAuI82kjCvxJ9mJuW1sV/5AzEfypfIRFJPwsRgFTBAXq6hNkcPsbo16Tx2+/MJ UDuuofUE00HmnIt2YfPVFJxrnSV5Z6ppaU7p0tujNEB3/2H0aWYGIKF1TvnEht+rXQef fb4gyQY3E93gMwe5REYDvIo2QMLH2QFb70Y2YfGRnpUEfdsDParp3FF9MNZlabX7v4Vr PLHQ==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=L2rribcOjnwt4VNORgwAor4H7h6UauEEJ+gLIsmop6Y=; b=gVRY+i90FPQmE2BmFq6WpboxHrwgy8lqLYUyKYgaSJIJ/eCRiYAwjScSIrQJwY09CE OOybwgGTDzqfsVdQvLcSa7LQgrnBGQyAprNPBnp99Sqv6ejHLDhgrMFA0rlsjdb8c8WX L2KOCC7LQhDBo6UxQKvdQil65aFpcNEQsOFB5IHeO+nz/1k2+TTQU0z7UPs2x18F6VFX RmVwq9zvCAzS8XJ/91OBDPQB5Y+tekZT/8iTae41wvBcy0J8o95ZKwPSXCGjwAzDDuU4 T29JRPRsQjOrul449I5vxN63MXutejv2z5qouHHWTyW+NG5pCqTnAjJeinHrX+jttlB/ L0Ig==
X-Gm-Message-State: AOAM532+Z6bC6a3QoClPzC7uDFJ5/DMVPfEqJYq7v/K497o+maF8LCcc 0vch+vu+PSs9j5YKtQRyQDRJVLfQzSSilIQSV1B8vA==
X-Google-Smtp-Source: ABdhPJych4O5BP7PfC28W/bG+1/a8LEZelB6Mkv9Vx7/2J2LCqBmNdGTOL6gEv0oohQ6aG/7INfuFQIqruwrpIDh//M=
X-Received: by 2002:a05:600c:4111:: with SMTP id j17mr9867781wmi.7.1642699177891;  Thu, 20 Jan 2022 09:19:37 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com>
In-Reply-To: <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com>
From: Phillipp Schoppmann <schoppmann@google.com>
Date: Thu, 20 Jan 2022 18:19:26 +0100
Message-ID: <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com>
To: Charlie Harrison <csharrison@google.com>
Cc: Christopher Patton <cpatton@cloudflare.com>, Erik Taubeneck <erik.taubeneck@gmail.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000f9550805d606b493"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/TLVjveJcDYmkYBoWWgNyGsv99tc>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 17:20:01 -0000

--000000000000f9550805d606b493
Content-Type: text/plain; charset="UTF-8"

Hi everyone,

I'm just reading through the discussion and catching up.
@Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
<csharrison@google.com> regarding the sizes of IDPF shares, I did some
benchmarks:

   - 32 bit inputs -> 2127 byte shares
   - 64 bit inputs -> 4283 byte shares
   - 128 bit inputs -> 8539 byte shares

These are for a DPF that allows evaluation at every bit of the input (like
Poplar), but currently without the VDAF part. I am in the process of
implementing verification for our IDPF implementation, and can tell you
more once that's done.

Best,
Phillipp


On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
wrote:

> Thanks Chris,
> I think I am in agreement with Erik that we could mitigate the issues with
> Option #2 if there is an ingestor client. I think framing the protocol this
> way actually allows close imitation of option #1 and #2
>
> - An "option 1" deployment could just specify one of the aggregators to
> work as the ingestor even if it isn't required by the protocol. They just
> happen to be the same party.
> - An "option 2" deployment could have the ingestor fan out reports to
> multiple aggregators (or ignore the ingestor completely)
>
> Your other points make perfect sense to me!
>
> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <cpatton@cloudflare.com>
> wrote:
>
>>
>>
>>
>>> Great! My initial read was that it wouldn't currently enable it, and I'm
>>> happy to work on an issue/PR that can make that more clear and enumerate
>>> various deployment scenarios, as you suggest. One point of confusion was
>>> the suggestion about limiting reports from a single IP address, as I'd
>>> expect an intermediary server between the client and leader/aggregators to
>>> likely have a fixed IP address.
>>>
>>
>> Awesome! Let's start with opening an issue and folks to start putting
>> their deployment setups there.
>>
>> Yes, rate-limiting clients based on IP alone is impossible for
>> aggregators if they're behind a proxy. It should be noted that this is a
>> heuristic only that an attacker may be able to defeat in many ways. To deal
>> with with Sybil attacks in general, we'll definitely have to do a bit more
>> anyway.
>>
>>
>> Chris P.
>>
>

--000000000000f9550805d606b493
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hi everyone,<div><br></div><div>I&#39;m just reading throu=
gh the discussion and catching up.=C2=A0</div><div><a class=3D"gmail_plusre=
ply" id=3D"plusReplyChip-0" href=3D"mailto:cpatton@cloudflare.com" tabindex=
=3D"-1">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plusreply" id=3D"plu=
sReplyChip-1" href=3D"mailto:csharrison@google.com" tabindex=3D"-1">@Charli=
e Harrison</a>=C2=A0regarding the sizes of IDPF shares, I did some benchmar=
ks:</div><div><ul><li>32 bit inputs -&gt; 2127 byte shares</li><li>64 bit i=
nputs -&gt; 4283 byte shares</li><li>128 bit inputs -&gt; 8539 byte shares<=
/li></ul></div><div>These are for a DPF that allows evaluation at every bit=
 of the input (like Poplar), but currently without the VDAF part. I am in t=
he process of implementing verification for our IDPF implementation, and ca=
n tell you more once that&#39;s done.</div><div><br></div><div>Best,</div><=
div>Phillipp</div><div><br></div></div><br><div class=3D"gmail_quote"><div =
dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Charlie Ha=
rrison &lt;<a href=3D"mailto:csharrison@google.com">csharrison@google.com</=
a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0p=
x 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><d=
iv dir=3D"ltr">Thanks Chris,<div>I think I am in agreement with Erik that w=
e could mitigate the issues with Option #2 if there is an ingestor client. =
I think framing the protocol this way actually allows close imitation=C2=A0=
of option #1 and #2</div><div><br></div><div>- An &quot;option 1&quot; depl=
oyment could=C2=A0just=C2=A0specify one of the aggregators to work as the i=
ngestor even if it isn&#39;t required by the protocol. They just happen to =
be the same party.</div><div>- An &quot;option 2&quot; deployment could hav=
e the ingestor fan out reports to multiple aggregators (or ignore the inges=
tor completely)</div><div><br></div><div>Your other points make perfect sen=
se to me!</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a h=
ref=3D"mailto:cpatton@cloudflare.com" target=3D"_blank">cpatton@cloudflare.=
com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"marg=
in:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1e=
x"><div dir=3D"ltr"><br>=C2=A0<div class=3D"gmail_quote"><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div></div><div>Great! My=
 initial read was that it wouldn&#39;t currently enable it, and I&#39;m hap=
py to work on an issue/PR that can make that more clear and enumerate vario=
us deployment=C2=A0scenarios, as you suggest. One point of confusion was th=
e suggestion about limiting reports from a single IP address, as I&#39;d ex=
pect an intermediary server between the client and leader/aggregators to li=
kely have a fixed IP address.</div></div></blockquote><div><br></div><div>A=
wesome! Let&#39;s start with opening an issue and folks to start putting th=
eir deployment setups there.</div><div><br></div><div>Yes, rate-limiting cl=
ients based on IP alone is impossible for aggregators if they&#39;re behind=
 a proxy. It should be noted that this is a heuristic only that an attacker=
 may be able to defeat in many ways. To deal with with Sybil attacks in gen=
eral, we&#39;ll definitely have to do a bit more anyway.<br></div>=C2=A0</d=
iv><div class=3D"gmail_quote"><br></div><div class=3D"gmail_quote">Chris P.=
<br></div></div>
</blockquote></div>
</blockquote></div>

--000000000000f9550805d606b493--


From nobody Thu Jan 20 09:25:28 2022
Return-Path: <schoppmann@google.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id F2BC23A19D6 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 09:25:26 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id TVlJbvsRGoEL for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 09:25:22 -0800 (PST)
Received: from mail-wm1-x331.google.com (mail-wm1-x331.google.com [IPv6:2a00:1450:4864:20::331]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 87B383A19D0 for <ppm@ietf.org>; Thu, 20 Jan 2022 09:25:22 -0800 (PST)
Received: by mail-wm1-x331.google.com with SMTP id e9-20020a05600c4e4900b0034d23cae3f0so15376336wmq.2 for <ppm@ietf.org>; Thu, 20 Jan 2022 09:25:22 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=An46Y/9vHp378h+GplQt/8pHkl7gY+miAljRCb4HhA0=; b=B7QoiByyaVKWp66XNlMd02vdNSJLFM7HDrLX3QlJuvFt5jUASfeY7+x9WzLcFCyFIb BTgrqtYTV2wPqoWPmIyP6cSahZraexAyCVBi4pNGXkWykJ9tcKIcP3qXLZJVqX2A1Afc 2QgYlFCKUHyOE/8IQjfk9unISvOT7T4r9K8paKvKzd3rdS3fa4s/suOvjPQUG9c8VNzf Do/pzK91I/qmKYStAw1ILAeG8doMcFfwxbS4pLJy6yYcZI7aQH0c6/asfZFVC2fh7Bkl t+al1kg10GyRAYJedVvykHqX8o97+A8E4dwylziZ88JbhOhLuVR3dFEb440l3MG1qwtG DReg==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=An46Y/9vHp378h+GplQt/8pHkl7gY+miAljRCb4HhA0=; b=VoaQsPb9Rk4mYAFYbfvgWssEbXBuwI/FSDU3I3Z5Ix2Pnx7fv30FkDFeuuJ1aS4Tqy iz/Bh/ygfJPy3IPmc91vhkIMgB8yG4jmRN7fTySgzidWBcW1sK59O2HH/GCelrcKoaOO cV4Uqsi8JmyijOEKzkHX/nuUXuR78A8eurTfC7jZWioLC3tc63mTb54URbVH/2QLNN4x hSJ4B9fpC2xUEithMDr8B61sUjrAPHaMDKlSqmEX0JibQT1zPJ0WwUptwhsbu6VTx3nx aLWl0/38Yh1aRFy58gaBarugsf31G+Q0wrVRgt+uUCTiHLkw6Lbvk3KEfhop0+047kmn HogA==
X-Gm-Message-State: AOAM532CPwqcrhm+x0UowUerJVXhCOlpxneSH9fB2lPcnKldFoyPqnZB W7FTicOeH7m0iU/ZPNHNVwmeNjLxPPS7x1K+OH1M+g==
X-Google-Smtp-Source: ABdhPJz/MAx4xnSS8IhvFchFZ3AUIOtHqiByPUfbeEDvNnTpya8JyvEe1rzVR9HJVfpbBHNvbcl2lShaD7+Zb2qJn+0=
X-Received: by 2002:adf:e949:: with SMTP id m9mr85998wrn.172.1642699519360; Thu, 20 Jan 2022 09:25:19 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com>
In-Reply-To: <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com>
From: Phillipp Schoppmann <schoppmann@google.com>
Date: Thu, 20 Jan 2022 18:25:07 +0100
Message-ID: <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com>
To: Charlie Harrison <csharrison@google.com>
Cc: Christopher Patton <cpatton@cloudflare.com>, Erik Taubeneck <erik.taubeneck@gmail.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000053b15f05d606c9ed"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/kZdBsZnJf0mTINoIY2BtZHeR5uo>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 17:25:27 -0000

--00000000000053b15f05d606c9ed
Content-Type: text/plain; charset="UTF-8"

One more note on the share sizes: A lot of overhead (~50%) in our
implementation comes from the way I used protocol buffers for serialization.
I think there is some optimization potential here, so the cost could be
about half that if we use a format that just sends the raw bytes needed.
For example, with 128 bit inputs, the theoretical lower bound of the cost
is just 4128 bytes (again, excluding the verification part).

On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <schoppmann@google.com>
wrote:

> Hi everyone,
>
> I'm just reading through the discussion and catching up.
> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
> <csharrison@google.com> regarding the sizes of IDPF shares, I did some
> benchmarks:
>
>    - 32 bit inputs -> 2127 byte shares
>    - 64 bit inputs -> 4283 byte shares
>    - 128 bit inputs -> 8539 byte shares
>
> These are for a DPF that allows evaluation at every bit of the input (like
> Poplar), but currently without the VDAF part. I am in the process of
> implementing verification for our IDPF implementation, and can tell you
> more once that's done.
>
> Best,
> Phillipp
>
>
> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
> wrote:
>
>> Thanks Chris,
>> I think I am in agreement with Erik that we could mitigate the issues
>> with Option #2 if there is an ingestor client. I think framing the protocol
>> this way actually allows close imitation of option #1 and #2
>>
>> - An "option 1" deployment could just specify one of the aggregators to
>> work as the ingestor even if it isn't required by the protocol. They just
>> happen to be the same party.
>> - An "option 2" deployment could have the ingestor fan out reports to
>> multiple aggregators (or ignore the ingestor completely)
>>
>> Your other points make perfect sense to me!
>>
>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>> cpatton@cloudflare.com> wrote:
>>
>>>
>>>
>>>
>>>> Great! My initial read was that it wouldn't currently enable it, and
>>>> I'm happy to work on an issue/PR that can make that more clear and
>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>> confusion was the suggestion about limiting reports from a single IP
>>>> address, as I'd expect an intermediary server between the client and
>>>> leader/aggregators to likely have a fixed IP address.
>>>>
>>>
>>> Awesome! Let's start with opening an issue and folks to start putting
>>> their deployment setups there.
>>>
>>> Yes, rate-limiting clients based on IP alone is impossible for
>>> aggregators if they're behind a proxy. It should be noted that this is a
>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>> anyway.
>>>
>>>
>>> Chris P.
>>>
>>

--00000000000053b15f05d606c9ed
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">One more note on the share sizes: A lot of overhead (~50%)=
 in our implementation comes from the way I used protocol buffers for seria=
lization.<div>I think there is some optimization potential here, so the cos=
t could be about half that if we use a format that just sends the raw bytes=
 needed.</div><div>For example, with 128 bit inputs, the theoretical lower =
bound of the cost is just 4128 bytes (again, excluding the verification par=
t).</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann &lt;<a href=3D=
"mailto:schoppmann@google.com">schoppmann@google.com</a>&gt; wrote:<br></di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Hi eve=
ryone,<div><br></div><div>I&#39;m just reading through the discussion and c=
atching up.=C2=A0</div><div><a class=3D"gmail_plusreply" id=3D"gmail-m_-724=
7445031704786523plusReplyChip-0" href=3D"mailto:cpatton@cloudflare.com" tar=
get=3D"_blank">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plusreply" id=
=3D"gmail-m_-7247445031704786523plusReplyChip-1" href=3D"mailto:csharrison@=
google.com" target=3D"_blank">@Charlie Harrison</a>=C2=A0regarding the size=
s of IDPF shares, I did some benchmarks:</div><div><ul><li>32 bit inputs -&=
gt; 2127 byte shares</li><li>64 bit inputs -&gt; 4283 byte shares</li><li>1=
28 bit inputs -&gt; 8539 byte shares</li></ul></div><div>These are for a DP=
F that allows evaluation at every bit of the input (like Poplar), but curre=
ntly without the VDAF part. I am in the process of implementing verificatio=
n for our IDPF implementation, and can tell you more once that&#39;s done.<=
/div><div><br></div><div>Best,</div><div>Phillipp</div><div><br></div></div=
><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Th=
u, Jan 20, 2022 at 5:04 PM Charlie Harrison &lt;<a href=3D"mailto:csharriso=
n@google.com" target=3D"_blank">csharrison@google.com</a>&gt; wrote:<br></d=
iv><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bord=
er-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Thank=
s Chris,<div>I think I am in agreement with Erik that we could mitigate the=
 issues with Option #2 if there is an ingestor client. I think framing the =
protocol this way actually allows close imitation=C2=A0of option #1 and #2<=
/div><div><br></div><div>- An &quot;option 1&quot; deployment could=C2=A0ju=
st=C2=A0specify one of the aggregators to work as the ingestor even if it i=
sn&#39;t required by the protocol. They just happen to be the same party.</=
div><div>- An &quot;option 2&quot; deployment could have the ingestor fan o=
ut reports to multiple aggregators (or ignore the ingestor completely)</div=
><div><br></div><div>Your other points make perfect sense to me!</div></div=
><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On We=
d, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a href=3D"mailto:cpatton=
@cloudflare.com" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br=
></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;=
border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><=
br>=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex"><div dir=3D"ltr"><div></div><div>Great! My initial read was that=
 it wouldn&#39;t currently enable it, and I&#39;m happy to work on an issue=
/PR that can make that more clear and enumerate various deployment=C2=A0sce=
narios, as you suggest. One point of confusion was the suggestion about lim=
iting reports from a single IP address, as I&#39;d expect an intermediary s=
erver between the client and leader/aggregators to likely have a fixed IP a=
ddress.</div></div></blockquote><div><br></div><div>Awesome! Let&#39;s star=
t with opening an issue and folks to start putting their deployment setups =
there.</div><div><br></div><div>Yes, rate-limiting clients based on IP alon=
e is impossible for aggregators if they&#39;re behind a proxy. It should be=
 noted that this is a heuristic only that an attacker may be able to defeat=
 in many ways. To deal with with Sybil attacks in general, we&#39;ll defini=
tely have to do a bit more anyway.<br></div>=C2=A0</div><div class=3D"gmail=
_quote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>

--00000000000053b15f05d606c9ed--


From nobody Thu Jan 20 11:19:03 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 5C3AD3A1D51 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 11:19:02 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.099
X-Spam-Level: 
X-Spam-Status: No, score=-7.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id tk4s0CfwkHU1 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 11:18:58 -0800 (PST)
Received: from mail-pl1-x62e.google.com (mail-pl1-x62e.google.com [IPv6:2607:f8b0:4864:20::62e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 081213A1D4B for <ppm@ietf.org>; Thu, 20 Jan 2022 11:18:57 -0800 (PST)
Received: by mail-pl1-x62e.google.com with SMTP id x11so198164plg.6 for <ppm@ietf.org>; Thu, 20 Jan 2022 11:18:57 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=VAdov1a+5BLn5S7kHsCMiIX0AqOLvlOAJUMfjNk7g80=; b=BWRF9lUuCoKKycCZ4/wYdiusa/g+7wgGjwAtHDUkOdpDrKP4+qf3fmVWQOtTP+3bKP tHE33cpqaVP5p+5csWU3bv+kfuHVwr4J7s2FnisAXfXCMSwvrYYN+i+PqCwpW1mUqOW+ zof+IJm6Eh0Jv/Hjf8Ukx7v5Z1C/jLtZJtu3Q=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=VAdov1a+5BLn5S7kHsCMiIX0AqOLvlOAJUMfjNk7g80=; b=st0tnuQdizD0wO8r87+CIdxbdoe9SmuoHpM20/reCvwF2MSwfxxR2puvJskw+4dVY4 hPbnerJ67coxZx8TfVa5BMSsCfTboAi3M79F4WS8ZIBXZqx6pe/PTg15oC3yHK1qifcv 6loO1o0IMABhFWNzTTPUfSGUpvQnLmBpGxOqcdrLuV0t/kQHrU/0w+R06nLmB+wWH2tt x4wxEEV6AO6agRCJj0iHCeD2j3TWqc/oGANWeZ8dBVCQUVxMuneBgqpM7HAXc+RUgbMV wSM93fDTGQ5ZO4xo5FFram1fNGdAcmCMQu8eCHTnlMOHVgtwO1QEacRAwXo5QI+57TzX tiuw==
X-Gm-Message-State: AOAM530iUJpCyxvDNJf0q3WJq/Gj3nQPoiiDzYdoWMJTlOGEcIOd5EGF zq2/5epAz3/6mLLmNIn4l3Wl6AmWaPD8JzzG8L+oyQ==
X-Google-Smtp-Source: ABdhPJy2caCIfZsVzmZHXH9nLqWdv95Lgy4Y9LieKS0Q17eIsK1GlSu57oGLnXWpBgviccK30fmpI2vd+il915hveiI=
X-Received: by 2002:a17:902:e202:b0:14a:e525:b675 with SMTP id u2-20020a170902e20200b0014ae525b675mr157879plb.11.1642706335315; Thu, 20 Jan 2022 11:18:55 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com>
In-Reply-To: <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Thu, 20 Jan 2022 11:18:44 -0800
Message-ID: <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com>
To: Phillipp Schoppmann <schoppmann@google.com>
Cc: Charlie Harrison <csharrison@google.com>, Erik Taubeneck <erik.taubeneck@gmail.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000009701b105d6085fa2"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/DHnFzkRT3SbQoQ60hpB4i9C7H_0>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 19:19:02 -0000

--0000000000009701b105d6085fa2
Content-Type: text/plain; charset="UTF-8"

Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
encoding for the input shares (including the IDPF key shares), which we can
reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
of data to re-transmit per share. If we go with Option #1, we'll definitely
have to figure out a way to mitigate this. (Option #4 is one possibility.)

I also agree with Erik and Charlie that Option #2's downsides can be
mitigated by putting an ingestor in front of the helper. Though it's
important to note that this probably cancels out the egress savings, since
the ingestor (itself a service provider) is proxying report shares to the
helper. To get the egress savings, the helper would have to be able to
absorb traffic from clients directly.

Either way, Option #2 will require extra protocol logic to deal with
synchronization of the report shares. The current spec has each aggregate
sub-request from the leader specify the nonce of the report being
aggregated. To respond, the helper will first check if it has a report
share with that nonce. If not (perhaps the report share hasn't arrived),
it'll have to treat this as an error. Where things get tricky is what the
leader does next:
1. It might retry later, but there's no guarantee that the helper's report
share will ever arrive. Eventually it'll have to give up and the report
will get dropped.
2. It might just ignore the report. If the helper's report share ever shows
up, it might end up storing it indefinitely, waiting for the leader to send
a request that never arrives.

This suggests that some data loss may be inevitable in some deployments of
Option #2. In my own opinion, this data loss is acceptable.

Chris P.






On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <schoppmann@google.com>
wrote:

> One more note on the share sizes: A lot of overhead (~50%) in our
> implementation comes from the way I used protocol buffers for serialization.
> I think there is some optimization potential here, so the cost could be
> about half that if we use a format that just sends the raw bytes needed.
> For example, with 128 bit inputs, the theoretical lower bound of the cost
> is just 4128 bytes (again, excluding the verification part).
>
> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <schoppmann@google.com>
> wrote:
>
>> Hi everyone,
>>
>> I'm just reading through the discussion and catching up.
>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>> <csharrison@google.com> regarding the sizes of IDPF shares, I did some
>> benchmarks:
>>
>>    - 32 bit inputs -> 2127 byte shares
>>    - 64 bit inputs -> 4283 byte shares
>>    - 128 bit inputs -> 8539 byte shares
>>
>> These are for a DPF that allows evaluation at every bit of the input
>> (like Poplar), but currently without the VDAF part. I am in the process of
>> implementing verification for our IDPF implementation, and can tell you
>> more once that's done.
>>
>> Best,
>> Phillipp
>>
>>
>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
>> wrote:
>>
>>> Thanks Chris,
>>> I think I am in agreement with Erik that we could mitigate the issues
>>> with Option #2 if there is an ingestor client. I think framing the protocol
>>> this way actually allows close imitation of option #1 and #2
>>>
>>> - An "option 1" deployment could just specify one of the aggregators to
>>> work as the ingestor even if it isn't required by the protocol. They just
>>> happen to be the same party.
>>> - An "option 2" deployment could have the ingestor fan out reports to
>>> multiple aggregators (or ignore the ingestor completely)
>>>
>>> Your other points make perfect sense to me!
>>>
>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>> cpatton@cloudflare.com> wrote:
>>>
>>>>
>>>>
>>>>
>>>>> Great! My initial read was that it wouldn't currently enable it, and
>>>>> I'm happy to work on an issue/PR that can make that more clear and
>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>> address, as I'd expect an intermediary server between the client and
>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>
>>>>
>>>> Awesome! Let's start with opening an issue and folks to start putting
>>>> their deployment setups there.
>>>>
>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>> anyway.
>>>>
>>>>
>>>> Chris P.
>>>>
>>>

--0000000000009701b105d6085fa2
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Thanks for the numbers, Philllipp! The VDAF spec will=
 specify a binary encoding for the input shares (including the IDPF key sha=
res), which we can reasonably expect to be much closer to the lower bound. =
Still, 4KB is a lot of data to re-transmit per share. If we go with Option =
#1, we&#39;ll definitely have to figure out a way to mitigate this. (Option=
 #4 is one possibility.)<br></div><div><br></div><div>I also agree with Eri=
k and Charlie that Option #2&#39;s downsides can be mitigated by putting an=
 ingestor in front of the helper. Though it&#39;s important to note that th=
is probably cancels out the egress savings, since the ingestor (itself a se=
rvice provider) is proxying report shares to the helper. To get the egress =
savings, the helper would have to be able to absorb traffic from clients di=
rectly.</div><div><br></div><div>Either way, Option #2 will require extra p=
rotocol logic to deal with synchronization of the report shares. The curren=
t spec has each aggregate sub-request from the leader specify the nonce of =
the report being aggregated. To respond, the helper will first check if it =
has a report share with that nonce. If not (perhaps the report share hasn&#=
39;t arrived), it&#39;ll have to treat this as an error. Where things get t=
ricky is what the leader does next:<br></div><div>1. It might retry later, =
but there&#39;s no guarantee that the helper&#39;s report share will ever a=
rrive. Eventually it&#39;ll have to give up and the report will get dropped=
.<br></div><div>2. It might just ignore the report. If the helper&#39;s rep=
ort share ever shows up, it might end up storing it indefinitely, waiting f=
or the leader to send a request that never arrives.</div><div><br></div><di=
v>This suggests that some data loss may be inevitable in some deployments o=
f Option #2. In my own opinion, this data loss is acceptable.<br></div><div=
><br></div><div>Chris P.<br></div><div><br></div><div><br></div><div><br></=
div><div><br></div><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr=
" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann =
&lt;<a href=3D"mailto:schoppmann@google.com" target=3D"_blank">schoppmann@g=
oogle.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex"><div dir=3D"ltr">One more note on the share sizes: A lot of over=
head (~50%) in our implementation comes from the way I used protocol buffer=
s for serialization.<div>I think there is some optimization potential here,=
 so the cost could be about half that if we use a format that just sends th=
e raw bytes needed.</div><div>For example, with 128 bit inputs, the theoret=
ical lower bound of the cost is just 4128 bytes (again, excluding the verif=
ication part).</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" c=
lass=3D"gmail_attr">On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann &lt=
;<a href=3D"mailto:schoppmann@google.com" target=3D"_blank">schoppmann@goog=
le.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"m=
argin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left=
:1ex"><div dir=3D"ltr">Hi everyone,<div><br></div><div>I&#39;m just reading=
 through the discussion and catching up.=C2=A0</div><div><a class=3D"gmail_=
plusreply" id=3D"gmail-m_-5860546362101609079gmail-m_-7075653027769385596gm=
ail-m_-7247445031704786523plusReplyChip-0" href=3D"mailto:cpatton@cloudflar=
e.com" target=3D"_blank">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plu=
sreply" id=3D"gmail-m_-5860546362101609079gmail-m_-7075653027769385596gmail=
-m_-7247445031704786523plusReplyChip-1" href=3D"mailto:csharrison@google.co=
m" target=3D"_blank">@Charlie Harrison</a>=C2=A0regarding the sizes of IDPF=
 shares, I did some benchmarks:</div><div><ul><li>32 bit inputs -&gt; 2127 =
byte shares</li><li>64 bit inputs -&gt; 4283 byte shares</li><li>128 bit in=
puts -&gt; 8539 byte shares</li></ul></div><div>These are for a DPF that al=
lows evaluation at every bit of the input (like Poplar), but currently with=
out the VDAF part. I am in the process of implementing verification for our=
 IDPF implementation, and can tell you more once that&#39;s done.</div><div=
><br></div><div>Best,</div><div>Phillipp</div><div><br></div></div><br><div=
 class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20=
, 2022 at 5:04 PM Charlie Harrison &lt;<a href=3D"mailto:csharrison@google.=
com" target=3D"_blank">csharrison@google.com</a>&gt; wrote:<br></div><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1=
px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Thanks Chris,<=
div>I think I am in agreement with Erik that we could mitigate the issues w=
ith Option #2 if there is an ingestor client. I think framing the protocol =
this way actually allows close imitation=C2=A0of option #1 and #2</div><div=
><br></div><div>- An &quot;option 1&quot; deployment could=C2=A0just=C2=A0s=
pecify one of the aggregators to work as the ingestor even if it isn&#39;t =
required by the protocol. They just happen to be the same party.</div><div>=
- An &quot;option 2&quot; deployment could have the ingestor fan out report=
s to multiple aggregators (or ignore the ingestor completely)</div><div><br=
></div><div>Your other points make perfect sense to me!</div></div><br><div=
 class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 19=
, 2022 at 7:28 PM Christopher Patton &lt;<a href=3D"mailto:cpatton@cloudfla=
re.com" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br></div><b=
lockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-le=
ft:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><br>=C2=A0=
<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margi=
n:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex=
"><div dir=3D"ltr"><div></div><div>Great! My initial read was that it would=
n&#39;t currently enable it, and I&#39;m happy to work on an issue/PR that =
can make that more clear and enumerate various deployment=C2=A0scenarios, a=
s you suggest. One point of confusion was the suggestion about limiting rep=
orts from a single IP address, as I&#39;d expect an intermediary server bet=
ween the client and leader/aggregators to likely have a fixed IP address.</=
div></div></blockquote><div><br></div><div>Awesome! Let&#39;s start with op=
ening an issue and folks to start putting their deployment setups there.</d=
iv><div><br></div><div>Yes, rate-limiting clients based on IP alone is impo=
ssible for aggregators if they&#39;re behind a proxy. It should be noted th=
at this is a heuristic only that an attacker may be able to defeat in many =
ways. To deal with with Sybil attacks in general, we&#39;ll definitely have=
 to do a bit more anyway.<br></div>=C2=A0</div><div class=3D"gmail_quote"><=
br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>

--0000000000009701b105d6085fa2--


From nobody Thu Jan 20 13:16:35 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 256FA3A2227 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:16:33 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.096
X-Spam-Level: 
X-Spam-Status: No, score=-2.096 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id gcBwnQK_QZ28 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:16:29 -0800 (PST)
Received: from mail-lf1-x12c.google.com (mail-lf1-x12c.google.com [IPv6:2a00:1450:4864:20::12c]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 7AE0E3A2228 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:16:29 -0800 (PST)
Received: by mail-lf1-x12c.google.com with SMTP id x11so26526774lfa.2 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:16:29 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=7TcXc++/tPvtkNb/EDMQsRpq9oeM98MpWmQeDYyv7BE=; b=Z8CqEj7BMNFn/pTYoxMXhjve8e1PhbXdl+JYpa+vw9A2BAoLpqRg2bv/KvYxCpl+2u ku26TKQTNm7MFqEhxQB3eYHRcz6sfgUFaCgg6usVi1W+yfz8PNJGGNnRZNlVIiDHhIIV rDlw4/NHXSn11rMZUQ9731QUDhrQ0XJatDmyoKCMzA8GmEwtMmZ2SBqa6dW/7jlz6AcY zEQK+UaF4Dht5DSdFSX2jn1cEN0hgpm5RpjMdo+EZ8cCO9b0hK6iIG98PqP74mskn9AE ne6vzDUVLK2Z5VqmSr8ZFGP6p+t44mZR7LDaW+f+0AuQx33Yj23kLDNVSjDxYVlEdixq dP9A==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=7TcXc++/tPvtkNb/EDMQsRpq9oeM98MpWmQeDYyv7BE=; b=0A6S+T62//8NVmaKvVFn45O+dPb3JdSwoXGksUfZ4BXqcOyS7JWaxANnfZuBxiyYEw sYhYJo65RP7TKanVGZzpu8nbByCDICoRS6G70BoVTkchU8MU5c4c+clXKOOXuIbgG0wc BEpTPFE4L4WigROgU4mP1jPGLnrWO9ebRsM4x91fmO9HbkLpa8hFjA8WZ+OwNeI2xwTX AoHjDAvY3pt13QpHgwIULdaQbHZZ9r7lV/5dglfkBRl8SN0tPHSKjuGFajJxEblwxii8 iRDFYVt8uaknHKWXc0ezhalFey/ZC1P4XTaJ+qW8xM2F2kMj2G6WrpM391kDs8RHTHXd 4a9Q==
X-Gm-Message-State: AOAM533Zzv4n5zNpmfOov1V/u0Lo2cSPBRm32fKZAfWZCIPnnm/6vBpH kIF7FFH5YABrr+m29SFJayoBpYsUyhXjVdtk5Nc=
X-Google-Smtp-Source: ABdhPJyUJYafPBprOtQn8FBB/8vddbnLJEwiu+Xx9WxamNi/PEPtXf+LPILlJvT4IrR6dOX7AOJKgEvgmwFF9uwzSrQ=
X-Received: by 2002:a2e:c51:: with SMTP id o17mr740289ljd.95.1642713386099; Thu, 20 Jan 2022 13:16:26 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com>
In-Reply-To: <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Thu, 20 Jan 2022 13:16:15 -0800
Message-ID: <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com>
To: Christopher Patton <cpatton@cloudflare.com>
Cc: Phillipp Schoppmann <schoppmann@google.com>, Charlie Harrison <csharrison@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000d8a90d05d60a0306"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/h6jqPSx_FCP8XasdIPjYmsLXYEQ>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 21:16:33 -0000

--000000000000d8a90d05d60a0306
Content-Type: text/plain; charset="UTF-8"

>
> To get the egress savings, the helper would have to be able to absorb
> traffic from clients directly.
>

I don't think this is always the case (though it depends on specific cloud
providers), but it would require the client to at least send the
various shares to different locations. Here is roughly what I envision:
helper1 (in cloudA)
helper2 (in cloudB)

ingestor.proxyA (in cloudA)
ingestor.proxyB (in cloudB)

If the client is able to send reports to ingestor.proxyA and
ingestor.proxyB, then ingestor.proxyA can forward reports to helper1, and
ingestor.proxyB can forward reports to helper2.  There should be ways in
many cloud providers for that forwarding to not incur egress fees (for
example, AWS allows that to happen with S3 so long as both accounts are
within the same region.)

I'm not sure if this is compatible with what you envision for Option #2,
but it does seem desirable for an ingestor to at least have this option for
cost reduction.

Erik

On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton <cpatton@cloudflare.com>
wrote:

> Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
> encoding for the input shares (including the IDPF key shares), which we can
> reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
> of data to re-transmit per share. If we go with Option #1, we'll definitely
> have to figure out a way to mitigate this. (Option #4 is one possibility.)
>
> I also agree with Erik and Charlie that Option #2's downsides can be
> mitigated by putting an ingestor in front of the helper. Though it's
> important to note that this probably cancels out the egress savings, since
> the ingestor (itself a service provider) is proxying report shares to the
> helper. To get the egress savings, the helper would have to be able to
> absorb traffic from clients directly.
>
> Either way, Option #2 will require extra protocol logic to deal with
> synchronization of the report shares. The current spec has each aggregate
> sub-request from the leader specify the nonce of the report being
> aggregated. To respond, the helper will first check if it has a report
> share with that nonce. If not (perhaps the report share hasn't arrived),
> it'll have to treat this as an error. Where things get tricky is what the
> leader does next:
> 1. It might retry later, but there's no guarantee that the helper's report
> share will ever arrive. Eventually it'll have to give up and the report
> will get dropped.
> 2. It might just ignore the report. If the helper's report share ever
> shows up, it might end up storing it indefinitely, waiting for the leader
> to send a request that never arrives.
>
> This suggests that some data loss may be inevitable in some deployments of
> Option #2. In my own opinion, this data loss is acceptable.
>
> Chris P.
>
>
>
>
>
>
> On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <schoppmann@google.com>
> wrote:
>
>> One more note on the share sizes: A lot of overhead (~50%) in our
>> implementation comes from the way I used protocol buffers for serialization.
>> I think there is some optimization potential here, so the cost could be
>> about half that if we use a format that just sends the raw bytes needed.
>> For example, with 128 bit inputs, the theoretical lower bound of the cost
>> is just 4128 bytes (again, excluding the verification part).
>>
>> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <
>> schoppmann@google.com> wrote:
>>
>>> Hi everyone,
>>>
>>> I'm just reading through the discussion and catching up.
>>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>>> <csharrison@google.com> regarding the sizes of IDPF shares, I did some
>>> benchmarks:
>>>
>>>    - 32 bit inputs -> 2127 byte shares
>>>    - 64 bit inputs -> 4283 byte shares
>>>    - 128 bit inputs -> 8539 byte shares
>>>
>>> These are for a DPF that allows evaluation at every bit of the input
>>> (like Poplar), but currently without the VDAF part. I am in the process of
>>> implementing verification for our IDPF implementation, and can tell you
>>> more once that's done.
>>>
>>> Best,
>>> Phillipp
>>>
>>>
>>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
>>> wrote:
>>>
>>>> Thanks Chris,
>>>> I think I am in agreement with Erik that we could mitigate the issues
>>>> with Option #2 if there is an ingestor client. I think framing the protocol
>>>> this way actually allows close imitation of option #1 and #2
>>>>
>>>> - An "option 1" deployment could just specify one of the aggregators to
>>>> work as the ingestor even if it isn't required by the protocol. They just
>>>> happen to be the same party.
>>>> - An "option 2" deployment could have the ingestor fan out reports to
>>>> multiple aggregators (or ignore the ingestor completely)
>>>>
>>>> Your other points make perfect sense to me!
>>>>
>>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>>> cpatton@cloudflare.com> wrote:
>>>>
>>>>>
>>>>>
>>>>>
>>>>>> Great! My initial read was that it wouldn't currently enable it, and
>>>>>> I'm happy to work on an issue/PR that can make that more clear and
>>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>>> address, as I'd expect an intermediary server between the client and
>>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>>
>>>>>
>>>>> Awesome! Let's start with opening an issue and folks to start putting
>>>>> their deployment setups there.
>>>>>
>>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>>> anyway.
>>>>>
>>>>>
>>>>> Chris P.
>>>>>
>>>>

--000000000000d8a90d05d60a0306
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">To get t=
he egress savings, the helper would have to be able to absorb traffic from =
clients directly.<br></blockquote><div><br></div><div>I don&#39;t think thi=
s is always the case (though=C2=A0it depends on specific cloud providers), =
but it would require the client to at least send the various=C2=A0shares to=
 different locations. Here is roughly what I envision:</div><div>helper1 (i=
n cloudA)</div><div>helper2 (in cloudB)</div><div><br></div><div>ingestor.p=
roxyA (in cloudA)</div><div>ingestor.proxyB (in cloudB)</div><div><br></div=
><div>If the client is able to send reports to ingestor.proxyA and ingestor=
.proxyB, then ingestor.proxyA can forward reports to helper1, and ingestor.=
proxyB can forward reports to helper2.=C2=A0 There should be ways in many c=
loud providers for that forwarding to not incur egress fees (for example, A=
WS allows that to happen with S3 so long as both accounts are within the sa=
me region.)</div><div><br></div><div>I&#39;m not sure if this is compatible=
 with what you envision for Option #2, but it does seem desirable for an in=
gestor to at least have this option for cost reduction.</div><div><br></div=
><div>Erik</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton &lt;<a =
href=3D"mailto:cpatton@cloudflare.com">cpatton@cloudflare.com</a>&gt; wrote=
:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.=
8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"lt=
r"><div>Thanks for the numbers, Philllipp! The VDAF spec will specify a bin=
ary encoding for the input shares (including the IDPF key shares), which we=
 can reasonably expect to be much closer to the lower bound. Still, 4KB is =
a lot of data to re-transmit per share. If we go with Option #1, we&#39;ll =
definitely have to figure out a way to mitigate this. (Option #4 is one pos=
sibility.)<br></div><div><br></div><div>I also agree with Erik and Charlie =
that Option #2&#39;s downsides can be mitigated by putting an ingestor in f=
ront of the helper. Though it&#39;s important to note that this probably ca=
ncels out the egress savings, since the ingestor (itself a service provider=
) is proxying report shares to the helper. To get the egress savings, the h=
elper would have to be able to absorb traffic from clients directly.</div><=
div><br></div><div>Either way, Option #2 will require extra protocol logic =
to deal with synchronization of the report shares. The current spec has eac=
h aggregate sub-request from the leader specify the nonce of the report bei=
ng aggregated. To respond, the helper will first check if it has a report s=
hare with that nonce. If not (perhaps the report share hasn&#39;t arrived),=
 it&#39;ll have to treat this as an error. Where things get tricky is what =
the leader does next:<br></div><div>1. It might retry later, but there&#39;=
s no guarantee that the helper&#39;s report share will ever arrive. Eventua=
lly it&#39;ll have to give up and the report will get dropped.<br></div><di=
v>2. It might just ignore the report. If the helper&#39;s report share ever=
 shows up, it might end up storing it indefinitely, waiting for the leader =
to send a request that never arrives.</div><div><br></div><div>This suggest=
s that some data loss may be inevitable in some deployments of Option #2. I=
n my own opinion, this data loss is acceptable.<br></div><div><br></div><di=
v>Chris P.<br></div><div><br></div><div><br></div><div><br></div><div><br><=
/div><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann &lt;<a href=3D=
"mailto:schoppmann@google.com" target=3D"_blank">schoppmann@google.com</a>&=
gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0=
px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div =
dir=3D"ltr">One more note on the share sizes: A lot of overhead (~50%) in o=
ur implementation comes from the way I used protocol buffers for serializat=
ion.<div>I think there is some optimization potential here, so the cost cou=
ld be about half that if we use a format that just sends the raw bytes need=
ed.</div><div>For example, with 128 bit inputs, the theoretical lower bound=
 of the cost is just 4128 bytes (again, excluding the verification part).</=
div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_at=
tr">On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann &lt;<a href=3D"mail=
to:schoppmann@google.com" target=3D"_blank">schoppmann@google.com</a>&gt; w=
rote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0p=
x 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=
=3D"ltr">Hi everyone,<div><br></div><div>I&#39;m just reading through the d=
iscussion and catching up.=C2=A0</div><div><a class=3D"gmail_plusreply" id=
=3D"gmail-m_-4044173416287307120gmail-m_-5860546362101609079gmail-m_-707565=
3027769385596gmail-m_-7247445031704786523plusReplyChip-0" href=3D"mailto:cp=
atton@cloudflare.com" target=3D"_blank">@Christopher Patton</a>=C2=A0<a cla=
ss=3D"gmail_plusreply" id=3D"gmail-m_-4044173416287307120gmail-m_-586054636=
2101609079gmail-m_-7075653027769385596gmail-m_-7247445031704786523plusReply=
Chip-1" href=3D"mailto:csharrison@google.com" target=3D"_blank">@Charlie Ha=
rrison</a>=C2=A0regarding the sizes of IDPF shares, I did some benchmarks:<=
/div><div><ul><li>32 bit inputs -&gt; 2127 byte shares</li><li>64 bit input=
s -&gt; 4283 byte shares</li><li>128 bit inputs -&gt; 8539 byte shares</li>=
</ul></div><div>These are for a DPF that allows evaluation at every bit of =
the input (like Poplar), but currently without the VDAF part. I am in the p=
rocess of implementing verification for our IDPF implementation, and can te=
ll you more once that&#39;s done.</div><div><br></div><div>Best,</div><div>=
Phillipp</div><div><br></div></div><br><div class=3D"gmail_quote"><div dir=
=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Charlie Harri=
son &lt;<a href=3D"mailto:csharrison@google.com" target=3D"_blank">csharris=
on@google.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div dir=3D"ltr">Thanks Chris,<div>I think I am in agreement w=
ith Erik that we could mitigate the issues with Option #2 if there is an in=
gestor client. I think framing the protocol this way actually allows close =
imitation=C2=A0of option #1 and #2</div><div><br></div><div>- An &quot;opti=
on 1&quot; deployment could=C2=A0just=C2=A0specify one of the aggregators t=
o work as the ingestor even if it isn&#39;t required by the protocol. They =
just happen to be the same party.</div><div>- An &quot;option 2&quot; deplo=
yment could have the ingestor fan out reports to multiple aggregators (or i=
gnore the ingestor completely)</div><div><br></div><div>Your other points m=
ake perfect sense to me!</div></div><br><div class=3D"gmail_quote"><div dir=
=3D"ltr" class=3D"gmail_attr">On Wed, Jan 19, 2022 at 7:28 PM Christopher P=
atton &lt;<a href=3D"mailto:cpatton@cloudflare.com" target=3D"_blank">cpatt=
on@cloudflare.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote"=
 style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);p=
adding-left:1ex"><div dir=3D"ltr"><br>=C2=A0<div class=3D"gmail_quote"><blo=
ckquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div></div><=
div>Great! My initial read was that it wouldn&#39;t currently enable it, an=
d I&#39;m happy to work on an issue/PR that can make that more clear and en=
umerate various deployment=C2=A0scenarios, as you suggest. One point of con=
fusion was the suggestion about limiting reports from a single IP address, =
as I&#39;d expect an intermediary server between the client and leader/aggr=
egators to likely have a fixed IP address.</div></div></blockquote><div><br=
></div><div>Awesome! Let&#39;s start with opening an issue and folks to sta=
rt putting their deployment setups there.</div><div><br></div><div>Yes, rat=
e-limiting clients based on IP alone is impossible for aggregators if they&=
#39;re behind a proxy. It should be noted that this is a heuristic only tha=
t an attacker may be able to defeat in many ways. To deal with with Sybil a=
ttacks in general, we&#39;ll definitely have to do a bit more anyway.<br></=
div>=C2=A0</div><div class=3D"gmail_quote"><br></div><div class=3D"gmail_qu=
ote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>

--000000000000d8a90d05d60a0306--


From nobody Thu Jan 20 13:25:04 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 6EF903A2264 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:25:02 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.099
X-Spam-Level: 
X-Spam-Status: No, score=-7.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id dBe0dH4wFD3J for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:24:57 -0800 (PST)
Received: from mail-pj1-x1032.google.com (mail-pj1-x1032.google.com [IPv6:2607:f8b0:4864:20::1032]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id B59D63A2263 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:24:57 -0800 (PST)
Received: by mail-pj1-x1032.google.com with SMTP id g9-20020a17090a67c900b001b4f1d71e4fso7052412pjm.4 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:24:57 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=0SayY3xLIzxYJcdDmxUbmu6S4zP5xM3REmiRa+LMu8Y=; b=Ob4kiv2WFfrkxRkAvr2qjmZS3XSENVxChjWESEc64g+jS/NmOs8K45D1AZgdWS0TyN e10namd0EPiT3uv3wRtYEMSm/wqpGYv1Y8M0EMqR13q6N9rOhzBFV+qDpdwEAcJkVb14 u3hILuxThS+ugSCLOqeSA8zIYTSpp3XPQIgwg=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=0SayY3xLIzxYJcdDmxUbmu6S4zP5xM3REmiRa+LMu8Y=; b=cyS3xOoOdy4uaNUBH93389fQJDpwomd/D3S1WI4eTOGEpnTHXH6ndND6i3sDA2fJbf 2ygOIvzrD2fkUvVINA6Cai5lozKGUVCjui8tEuQhwg/HMAD5UXjdDU5cfX73gROe1rjD ylrUWzFC+rZw5RBsncCBZsorJus5uLnvYaPhMz9CVJ43v90RA0hxvoeUR67e2yx65R8p vuBZ40L+V+utJyNZWaw1ZXJIakRwn4SYMNBriOlm+YmC6eagtI6nQdbLDbdhenFPujny 3rvD8IsNtXAIYKtrRKOGKtyoOooOUtslEBs3nR8gpRtTe1XWFz54/CoZscGSxedKCo8m 633Q==
X-Gm-Message-State: AOAM531ObA19fcmq+cUjno38XgtW5tCozOgZ0fpeWrJmh2w49VNgZrsN Hts1nAukXf6e+GVKT5ykQokSzMTJJwqW7wuaswIbFk22j1c=
X-Google-Smtp-Source: ABdhPJwcNUOR+jJBS1bAYJegVkAOp49YM6gEUZPKxjW3bCJ6xi1dAtxrEveYEB4aA5s4Ag2eol8tl2PIo5WGzrjosVA=
X-Received: by 2002:a17:902:c713:b0:14b:d31:f9cd with SMTP id p19-20020a170902c71300b0014b0d31f9cdmr896365plp.29.1642713896202; Thu, 20 Jan 2022 13:24:56 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com>
In-Reply-To: <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Thu, 20 Jan 2022 13:24:45 -0800
Message-ID: <CAG2Zi21W7jiu=TG1qOiSBbXAzwbTCT0TpD7OxFq7AWOa=ST13Q@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Phillipp Schoppmann <schoppmann@google.com>, Charlie Harrison <csharrison@google.com>,  Robert Kubis <rkubis@google.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000040c41205d60a22a8"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/_04JFAZiSYBdT0u50kmQhfLETYY>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 21:25:02 -0000

--00000000000040c41205d60a22a8
Content-Type: text/plain; charset="UTF-8"

Ohhh yeah that makes sense! Erik's setup seems compatible with Option #2.
The ingestor proxies would still have to speak to each other to coordinate,
but I buy that you could handle the synchronization while still reducing
egress cost. But again, not every deployment will look like this, so there
will still need to be a way in the core protocol to deal with
synchronization.

It sounds like most who have chimed in on this thread are in favor of
Option #2. I'd like to hear from those who object or have reservations.

Chris P.

On Thu, Jan 20, 2022 at 1:16 PM Erik Taubeneck <erik.taubeneck@gmail.com>
wrote:

> To get the egress savings, the helper would have to be able to absorb
>> traffic from clients directly.
>>
>
> I don't think this is always the case (though it depends on specific cloud
> providers), but it would require the client to at least send the
> various shares to different locations. Here is roughly what I envision:
> helper1 (in cloudA)
> helper2 (in cloudB)
>
> ingestor.proxyA (in cloudA)
> ingestor.proxyB (in cloudB)
>
> If the client is able to send reports to ingestor.proxyA and
> ingestor.proxyB, then ingestor.proxyA can forward reports to helper1, and
> ingestor.proxyB can forward reports to helper2.  There should be ways in
> many cloud providers for that forwarding to not incur egress fees (for
> example, AWS allows that to happen with S3 so long as both accounts are
> within the same region.)
>
> I'm not sure if this is compatible with what you envision for Option #2,
> but it does seem desirable for an ingestor to at least have this option for
> cost reduction.
>
> Erik
>
> On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton <
> cpatton@cloudflare.com> wrote:
>
>> Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
>> encoding for the input shares (including the IDPF key shares), which we can
>> reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
>> of data to re-transmit per share. If we go with Option #1, we'll definitely
>> have to figure out a way to mitigate this. (Option #4 is one possibility.)
>>
>> I also agree with Erik and Charlie that Option #2's downsides can be
>> mitigated by putting an ingestor in front of the helper. Though it's
>> important to note that this probably cancels out the egress savings, since
>> the ingestor (itself a service provider) is proxying report shares to the
>> helper. To get the egress savings, the helper would have to be able to
>> absorb traffic from clients directly.
>>
>> Either way, Option #2 will require extra protocol logic to deal with
>> synchronization of the report shares. The current spec has each aggregate
>> sub-request from the leader specify the nonce of the report being
>> aggregated. To respond, the helper will first check if it has a report
>> share with that nonce. If not (perhaps the report share hasn't arrived),
>> it'll have to treat this as an error. Where things get tricky is what the
>> leader does next:
>> 1. It might retry later, but there's no guarantee that the helper's
>> report share will ever arrive. Eventually it'll have to give up and the
>> report will get dropped.
>> 2. It might just ignore the report. If the helper's report share ever
>> shows up, it might end up storing it indefinitely, waiting for the leader
>> to send a request that never arrives.
>>
>> This suggests that some data loss may be inevitable in some deployments
>> of Option #2. In my own opinion, this data loss is acceptable.
>>
>> Chris P.
>>
>>
>>
>>
>>
>>
>> On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <
>> schoppmann@google.com> wrote:
>>
>>> One more note on the share sizes: A lot of overhead (~50%) in our
>>> implementation comes from the way I used protocol buffers for serialization.
>>> I think there is some optimization potential here, so the cost could be
>>> about half that if we use a format that just sends the raw bytes needed.
>>> For example, with 128 bit inputs, the theoretical lower bound of the
>>> cost is just 4128 bytes (again, excluding the verification part).
>>>
>>> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <
>>> schoppmann@google.com> wrote:
>>>
>>>> Hi everyone,
>>>>
>>>> I'm just reading through the discussion and catching up.
>>>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>>>> <csharrison@google.com> regarding the sizes of IDPF shares, I did some
>>>> benchmarks:
>>>>
>>>>    - 32 bit inputs -> 2127 byte shares
>>>>    - 64 bit inputs -> 4283 byte shares
>>>>    - 128 bit inputs -> 8539 byte shares
>>>>
>>>> These are for a DPF that allows evaluation at every bit of the input
>>>> (like Poplar), but currently without the VDAF part. I am in the process of
>>>> implementing verification for our IDPF implementation, and can tell you
>>>> more once that's done.
>>>>
>>>> Best,
>>>> Phillipp
>>>>
>>>>
>>>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
>>>> wrote:
>>>>
>>>>> Thanks Chris,
>>>>> I think I am in agreement with Erik that we could mitigate the issues
>>>>> with Option #2 if there is an ingestor client. I think framing the protocol
>>>>> this way actually allows close imitation of option #1 and #2
>>>>>
>>>>> - An "option 1" deployment could just specify one of the aggregators
>>>>> to work as the ingestor even if it isn't required by the protocol. They
>>>>> just happen to be the same party.
>>>>> - An "option 2" deployment could have the ingestor fan out reports to
>>>>> multiple aggregators (or ignore the ingestor completely)
>>>>>
>>>>> Your other points make perfect sense to me!
>>>>>
>>>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>>>> cpatton@cloudflare.com> wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> Great! My initial read was that it wouldn't currently enable it, and
>>>>>>> I'm happy to work on an issue/PR that can make that more clear and
>>>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>>>> address, as I'd expect an intermediary server between the client and
>>>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>>>
>>>>>>
>>>>>> Awesome! Let's start with opening an issue and folks to start putting
>>>>>> their deployment setups there.
>>>>>>
>>>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>>>> anyway.
>>>>>>
>>>>>>
>>>>>> Chris P.
>>>>>>
>>>>>

--00000000000040c41205d60a22a8
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Ohhh yeah that makes sense! Erik&#39;s setup seems co=
mpatible with Option #2. The ingestor proxies would still have to speak to =
each other to coordinate, but I buy that you could handle the synchronizati=
on while still reducing egress cost. But again, not every deployment will l=
ook like this, so there will still need to be a way in the core protocol to=
 deal with synchronization.</div><div><br></div><div>It sounds like most wh=
o have chimed in on this thread are in favor of Option #2. I&#39;d like to =
hear from those who object or have reservations.</div><div><br></div><div>C=
hris P.<br></div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" clas=
s=3D"gmail_attr">On Thu, Jan 20, 2022 at 1:16 PM Erik Taubeneck &lt;<a href=
=3D"mailto:erik.taubeneck@gmail.com" target=3D"_blank">erik.taubeneck@gmail=
.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"mar=
gin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1=
ex"><div dir=3D"ltr"><blockquote class=3D"gmail_quote" style=3D"margin:0px =
0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">To g=
et the egress savings, the helper would have to be able to absorb traffic f=
rom clients directly.<br></blockquote><div><br></div><div>I don&#39;t think=
 this is always the case (though=C2=A0it depends on specific cloud provider=
s), but it would require the client to at least send the various=C2=A0share=
s to different locations. Here is roughly what I envision:</div><div>helper=
1 (in cloudA)</div><div>helper2 (in cloudB)</div><div><br></div><div>ingest=
or.proxyA (in cloudA)</div><div>ingestor.proxyB (in cloudB)</div><div><br><=
/div><div>If the client is able to send reports to ingestor.proxyA and inge=
stor.proxyB, then ingestor.proxyA can forward reports to helper1, and inges=
tor.proxyB can forward reports to helper2.=C2=A0 There should be ways in ma=
ny cloud providers for that forwarding to not incur egress fees (for exampl=
e, AWS allows that to happen with S3 so long as both accounts are within th=
e same region.)</div><div><br></div><div>I&#39;m not sure if this is compat=
ible with what you envision for Option #2, but it does seem desirable for a=
n ingestor to at least have this option for cost reduction.</div><div><br><=
/div><div>Erik</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" c=
lass=3D"gmail_attr">On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton &lt=
;<a href=3D"mailto:cpatton@cloudflare.com" target=3D"_blank">cpatton@cloudf=
lare.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D=
"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-le=
ft:1ex"><div dir=3D"ltr"><div>Thanks for the numbers, Philllipp! The VDAF s=
pec will specify a binary encoding for the input shares (including the IDPF=
 key shares), which we can reasonably expect to be much closer to the lower=
 bound. Still, 4KB is a lot of data to re-transmit per share. If we go with=
 Option #1, we&#39;ll definitely have to figure out a way to mitigate this.=
 (Option #4 is one possibility.)<br></div><div><br></div><div>I also agree =
with Erik and Charlie that Option #2&#39;s downsides can be mitigated by pu=
tting an ingestor in front of the helper. Though it&#39;s important to note=
 that this probably cancels out the egress savings, since the ingestor (its=
elf a service provider) is proxying report shares to the helper. To get the=
 egress savings, the helper would have to be able to absorb traffic from cl=
ients directly.</div><div><br></div><div>Either way, Option #2 will require=
 extra protocol logic to deal with synchronization of the report shares. Th=
e current spec has each aggregate sub-request from the leader specify the n=
once of the report being aggregated. To respond, the helper will first chec=
k if it has a report share with that nonce. If not (perhaps the report shar=
e hasn&#39;t arrived), it&#39;ll have to treat this as an error. Where thin=
gs get tricky is what the leader does next:<br></div><div>1. It might retry=
 later, but there&#39;s no guarantee that the helper&#39;s report share wil=
l ever arrive. Eventually it&#39;ll have to give up and the report will get=
 dropped.<br></div><div>2. It might just ignore the report. If the helper&#=
39;s report share ever shows up, it might end up storing it indefinitely, w=
aiting for the leader to send a request that never arrives.</div><div><br><=
/div><div>This suggests that some data loss may be inevitable in some deplo=
yments of Option #2. In my own opinion, this data loss is acceptable.<br></=
div><div><br></div><div>Chris P.<br></div><div><br></div><div><br></div><di=
v><br></div><div><br></div><br></div><br><div class=3D"gmail_quote"><div di=
r=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 9:25 AM Phillipp Sch=
oppmann &lt;<a href=3D"mailto:schoppmann@google.com" target=3D"_blank">scho=
ppmann@google.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote"=
 style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);p=
adding-left:1ex"><div dir=3D"ltr">One more note on the share sizes: A lot o=
f overhead (~50%) in our implementation comes from the way I used protocol =
buffers for serialization.<div>I think there is some optimization potential=
 here, so the cost could be about half that if we use a format that just se=
nds the raw bytes needed.</div><div>For example, with 128 bit inputs, the t=
heoretical lower bound of the cost is just 4128 bytes (again, excluding the=
 verification part).</div></div><br><div class=3D"gmail_quote"><div dir=3D"=
ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppma=
nn &lt;<a href=3D"mailto:schoppmann@google.com" target=3D"_blank">schoppman=
n@google.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" styl=
e=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddin=
g-left:1ex"><div dir=3D"ltr">Hi everyone,<div><br></div><div>I&#39;m just r=
eading through the discussion and catching up.=C2=A0</div><div><a class=3D"=
gmail_plusreply" id=3D"gmail-m_-9124873169761746444gmail-m_6509964474706063=
333gmail-m_4775907138701339528gmail-m_3298218662031834129gmail-m_-404417341=
6287307120gmail-m_-5860546362101609079gmail-m_-7075653027769385596gmail-m_-=
7247445031704786523plusReplyChip-0" href=3D"mailto:cpatton@cloudflare.com" =
target=3D"_blank">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plusreply"=
 id=3D"gmail-m_-9124873169761746444gmail-m_6509964474706063333gmail-m_47759=
07138701339528gmail-m_3298218662031834129gmail-m_-4044173416287307120gmail-=
m_-5860546362101609079gmail-m_-7075653027769385596gmail-m_-7247445031704786=
523plusReplyChip-1" href=3D"mailto:csharrison@google.com" target=3D"_blank"=
>@Charlie Harrison</a>=C2=A0regarding the sizes of IDPF shares, I did some =
benchmarks:</div><div><ul><li>32 bit inputs -&gt; 2127 byte shares</li><li>=
64 bit inputs -&gt; 4283 byte shares</li><li>128 bit inputs -&gt; 8539 byte=
 shares</li></ul></div><div>These are for a DPF that allows evaluation at e=
very bit of the input (like Poplar), but currently without the VDAF part. I=
 am in the process of implementing verification for our IDPF implementation=
, and can tell you more once that&#39;s done.</div><div><br></div><div>Best=
,</div><div>Phillipp</div><div><br></div></div><br><div class=3D"gmail_quot=
e"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Ch=
arlie Harrison &lt;<a href=3D"mailto:csharrison@google.com" target=3D"_blan=
k">csharrison@google.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex"><div dir=3D"ltr">Thanks Chris,<div>I think I am in =
agreement with Erik that we could mitigate the issues with Option #2 if the=
re is an ingestor client. I think framing the protocol this way actually al=
lows close imitation=C2=A0of option #1 and #2</div><div><br></div><div>- An=
 &quot;option 1&quot; deployment could=C2=A0just=C2=A0specify one of the ag=
gregators to work as the ingestor even if it isn&#39;t required by the prot=
ocol. They just happen to be the same party.</div><div>- An &quot;option 2&=
quot; deployment could have the ingestor fan out reports to multiple aggreg=
ators (or ignore the ingestor completely)</div><div><br></div><div>Your oth=
er points make perfect sense to me!</div></div><br><div class=3D"gmail_quot=
e"><div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 19, 2022 at 7:28 PM Ch=
ristopher Patton &lt;<a href=3D"mailto:cpatton@cloudflare.com" target=3D"_b=
lank">cpatton@cloudflare.com</a>&gt; wrote:<br></div><blockquote class=3D"g=
mail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204=
,204,204);padding-left:1ex"><div dir=3D"ltr"><br>=C2=A0<div class=3D"gmail_=
quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;=
border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><=
div></div><div>Great! My initial read was that it wouldn&#39;t currently en=
able it, and I&#39;m happy to work on an issue/PR that can make that more c=
lear and enumerate various deployment=C2=A0scenarios, as you suggest. One p=
oint of confusion was the suggestion about limiting reports from a single I=
P address, as I&#39;d expect an intermediary server between the client and =
leader/aggregators to likely have a fixed IP address.</div></div></blockquo=
te><div><br></div><div>Awesome! Let&#39;s start with opening an issue and f=
olks to start putting their deployment setups there.</div><div><br></div><d=
iv>Yes, rate-limiting clients based on IP alone is impossible for aggregato=
rs if they&#39;re behind a proxy. It should be noted that this is a heurist=
ic only that an attacker may be able to defeat in many ways. To deal with w=
ith Sybil attacks in general, we&#39;ll definitely have to do a bit more an=
yway.<br></div>=C2=A0</div><div class=3D"gmail_quote"><br></div><div class=
=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>

--00000000000040c41205d60a22a8--


From nobody Thu Jan 20 13:37:06 2022
Return-Path: <watsonbladd@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 211613A22D2 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:37:04 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id DthiRxBBUhUt for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:37:01 -0800 (PST)
Received: from mail-ed1-x52f.google.com (mail-ed1-x52f.google.com [IPv6:2a00:1450:4864:20::52f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id BD4483A22D0 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:37:00 -0800 (PST)
Received: by mail-ed1-x52f.google.com with SMTP id r10so3352474edt.1 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:37:00 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=yK2Qwf9hAR9yODFp3plexs5A4TLU7eU+O34VJLAxKrQ=; b=Py0XbEisn7GRylPZCrhpmHd23tZsp3CMKiLak2i/LEqeZXldZ2ig64l2IrAx6n/8DZ mDtD5BHUEwjJTLsov3wgp7DZfoLbidTZQGXRu6Ydfss16m9lOVSJy3W0UvQ52tX64ebg xhTrlqWJGxYhUeD4tydDVwQqpmt88ziP7bRb66iOwJHQTQ8pqoE++no5LJSU3Xg4zbnN CBakf9BxiYn2iOxXNVJqWxPXa97+d+BT2G5RLhB/lQwi/PQegnzWerFzhWLQN1YAbiwg CM7MbWEOdwycFEOCyUUNP7jrYO6/ZHzaX6PD6Xn14a1lyOSEsS8T25GTWzJWz6Pzma7l Fb4w==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=yK2Qwf9hAR9yODFp3plexs5A4TLU7eU+O34VJLAxKrQ=; b=8NrRASjDqWliHlZIH9JmCV4aWV4hmZVSgZ49cfnEO1bL4lU5Gs8l8MqDfHz+QrcoYB RvJZ93+hGN3BIg1R2raCMoaVwCi5r1Owa6lABOak6ETN6GsDk+Jez7zqqQd/gdFtGe5B hFCe85gMKYT+lAqVPu6tf5BznU0IgIw/R3ixv+idtBW1LvopPRwa8jhOQBd4UadtNR/W uARm9e6K26zslxHJFasD6OxUXphNiHbZWPkBS1xQ4j+adp1WaHANbVRPp82qg1V4DxQ4 zGaxGrKC103zg3fJtpOm9JQkK08IEGcDIzx0swyv4oeSqMcD7PpH3454YT5o/zJW0q7q UTFA==
X-Gm-Message-State: AOAM531digRpciTUCBEfM2JU0astuoE3qTXEjl0o870vZuUGozqjaKsg UTfRz23eylt5tb5ZD7lX5lQ7/oXyFD4MrZvs8Cm0vXZo
X-Google-Smtp-Source: ABdhPJzG7olXBA8mPGAlSIEUBePaHfa/gDWWEEwncyxctrDHwGQs42xhe0vWrZo4gKxZ+IGlHvkqAFCO4gQW6qBD3TM=
X-Received: by 2002:a05:6402:51cd:: with SMTP id r13mr1043232edd.139.1642714617984;  Thu, 20 Jan 2022 13:36:57 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com>
In-Reply-To: <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com>
From: Watson Ladd <watsonbladd@gmail.com>
Date: Thu, 20 Jan 2022 16:36:48 -0500
Message-ID: <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Christopher Patton <cpatton@cloudflare.com>, Charlie Harrison <csharrison@google.com>,  Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>,  ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000045b9a805d60a4dc4"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/3mPex2u-l4e7L4O-RMqPhr5PNjw>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 21:37:04 -0000

--00000000000045b9a805d60a4dc4
Content-Type: text/plain; charset="UTF-8"

On Thu, Jan 20, 2022, 4:16 PM Erik Taubeneck <erik.taubeneck@gmail.com>
wrote:

> To get the egress savings, the helper would have to be able to absorb
>> traffic from clients directly.
>>
>
> I don't think this is always the case (though it depends on specific cloud
> providers), but it would require the client to at least send the
> various shares to different locations. Here is roughly what I envision:
> helper1 (in cloudA)
> helper2 (in cloudB)
>
> ingestor.proxyA (in cloudA)
> ingestor.proxyB (in cloudB)
>
> If the client is able to send reports to ingestor.proxyA and
> ingestor.proxyB, then ingestor.proxyA can forward reports to helper1, and
> ingestor.proxyB can forward reports to helper2.  There should be ways in
> many cloud providers for that forwarding to not incur egress fees (for
> example, AWS allows that to happen with S3 so long as both accounts are
> within the same region.)
>

Clients to ingestor replication creates many of the same challenges that
we're trying to solve with the ingestor, in particular partial availability
and replay if I'm understanding the proposal properly. At this point it's
not clear why the ingestor and aggregator need to be separate.

>
> I'm not sure if this is compatible with what you envision for Option #2,
> but it does seem desirable for an ingestor to at least have this option for
> cost reduction.
>

> Erik
>
> On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton <
> cpatton@cloudflare.com> wrote:
>
>> Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
>> encoding for the input shares (including the IDPF key shares), which we can
>> reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
>> of data to re-transmit per share. If we go with Option #1, we'll definitely
>> have to figure out a way to mitigate this. (Option #4 is one possibility.)
>>
>> I also agree with Erik and Charlie that Option #2's downsides can be
>> mitigated by putting an ingestor in front of the helper. Though it's
>> important to note that this probably cancels out the egress savings, since
>> the ingestor (itself a service provider) is proxying report shares to the
>> helper. To get the egress savings, the helper would have to be able to
>> absorb traffic from clients directly.
>>
>> Either way, Option #2 will require extra protocol logic to deal with
>> synchronization of the report shares. The current spec has each aggregate
>> sub-request from the leader specify the nonce of the report being
>> aggregated. To respond, the helper will first check if it has a report
>> share with that nonce. If not (perhaps the report share hasn't arrived),
>> it'll have to treat this as an error. Where things get tricky is what the
>> leader does next:
>> 1. It might retry later, but there's no guarantee that the helper's
>> report share will ever arrive. Eventually it'll have to give up and the
>> report will get dropped.
>> 2. It might just ignore the report. If the helper's report share ever
>> shows up, it might end up storing it indefinitely, waiting for the leader
>> to send a request that never arrives.
>>
>> This suggests that some data loss may be inevitable in some deployments
>> of Option #2. In my own opinion, this data loss is acceptable.
>>
>> Chris P.
>>
>>
>>
>>
>>
>>
>> On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <
>> schoppmann@google.com> wrote:
>>
>>> One more note on the share sizes: A lot of overhead (~50%) in our
>>> implementation comes from the way I used protocol buffers for serialization.
>>> I think there is some optimization potential here, so the cost could be
>>> about half that if we use a format that just sends the raw bytes needed.
>>> For example, with 128 bit inputs, the theoretical lower bound of the
>>> cost is just 4128 bytes (again, excluding the verification part).
>>>
>>> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <
>>> schoppmann@google.com> wrote:
>>>
>>>> Hi everyone,
>>>>
>>>> I'm just reading through the discussion and catching up.
>>>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>>>> <csharrison@google.com> regarding the sizes of IDPF shares, I did some
>>>> benchmarks:
>>>>
>>>>    - 32 bit inputs -> 2127 byte shares
>>>>    - 64 bit inputs -> 4283 byte shares
>>>>    - 128 bit inputs -> 8539 byte shares
>>>>
>>>> These are for a DPF that allows evaluation at every bit of the input
>>>> (like Poplar), but currently without the VDAF part. I am in the process of
>>>> implementing verification for our IDPF implementation, and can tell you
>>>> more once that's done.
>>>>
>>>> Best,
>>>> Phillipp
>>>>
>>>>
>>>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <csharrison@google.com>
>>>> wrote:
>>>>
>>>>> Thanks Chris,
>>>>> I think I am in agreement with Erik that we could mitigate the issues
>>>>> with Option #2 if there is an ingestor client. I think framing the protocol
>>>>> this way actually allows close imitation of option #1 and #2
>>>>>
>>>>> - An "option 1" deployment could just specify one of the aggregators
>>>>> to work as the ingestor even if it isn't required by the protocol. They
>>>>> just happen to be the same party.
>>>>> - An "option 2" deployment could have the ingestor fan out reports to
>>>>> multiple aggregators (or ignore the ingestor completely)
>>>>>
>>>>> Your other points make perfect sense to me!
>>>>>
>>>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>>>> cpatton@cloudflare.com> wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> Great! My initial read was that it wouldn't currently enable it, and
>>>>>>> I'm happy to work on an issue/PR that can make that more clear and
>>>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>>>> address, as I'd expect an intermediary server between the client and
>>>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>>>
>>>>>>
>>>>>> Awesome! Let's start with opening an issue and folks to start putting
>>>>>> their deployment setups there.
>>>>>>
>>>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>>>> anyway.
>>>>>>
>>>>>>
>>>>>> Chris P.
>>>>>>
>>>>> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--00000000000045b9a805d60a4dc4
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"auto"><div><br><br><div class=3D"gmail_quote"><div dir=3D"ltr" =
class=3D"gmail_attr">On Thu, Jan 20, 2022, 4:16 PM Erik Taubeneck &lt;<a hr=
ef=3D"mailto:erik.taubeneck@gmail.com" rel=3D"noreferrer noreferrer norefer=
rer noreferrer" target=3D"_blank">erik.taubeneck@gmail.com</a>&gt; wrote:<b=
r></div><blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border=
-left:1px #ccc solid;padding-left:1ex"><div dir=3D"ltr"><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex">To get the egress savings, the helper woul=
d have to be able to absorb traffic from clients directly.<br></blockquote>=
<div><br></div><div>I don&#39;t think this is always the case (though=C2=A0=
it depends on specific cloud providers), but it would require the client to=
 at least send the various=C2=A0shares to different locations. Here is roug=
hly what I envision:</div><div>helper1 (in cloudA)</div><div>helper2 (in cl=
oudB)</div><div><br></div><div>ingestor.proxyA (in cloudA)</div><div>ingest=
or.proxyB (in cloudB)</div><div><br></div><div>If the client is able to sen=
d reports to ingestor.proxyA and ingestor.proxyB, then ingestor.proxyA can =
forward reports to helper1, and ingestor.proxyB can forward reports to help=
er2.=C2=A0 There should be ways in many cloud providers for that forwarding=
 to not incur egress fees (for example, AWS allows that to happen with S3 s=
o long as both accounts are within the same region.)</div></div></blockquot=
e></div></div><div dir=3D"auto"><br></div><div dir=3D"auto">Clients to inge=
stor replication creates many of the same challenges that we&#39;re trying =
to solve with the ingestor, in particular partial availability and replay i=
f I&#39;m understanding the proposal properly. At this point it&#39;s not c=
lear why the ingestor and aggregator need to be separate.</div><div dir=3D"=
auto"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D=
"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir=3D=
"ltr"><div><br></div><div>I&#39;m not sure if this is compatible with what =
you envision for Option #2, but it does seem desirable for an ingestor to a=
t least have this option for cost reduction.</div></div></blockquote></div>=
</div><div dir=3D"auto"><div class=3D"gmail_quote"><blockquote class=3D"gma=
il_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-lef=
t:1ex"><div dir=3D"ltr"><div><br></div><div>Erik</div></div><br><div class=
=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022=
 at 11:18 AM Christopher Patton &lt;<a href=3D"mailto:cpatton@cloudflare.co=
m" rel=3D"noreferrer noreferrer noreferrer noreferrer noreferrer" target=3D=
"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br></div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>Thanks for the numbe=
rs, Philllipp! The VDAF spec will specify a binary encoding for the input s=
hares (including the IDPF key shares), which we can reasonably expect to be=
 much closer to the lower bound. Still, 4KB is a lot of data to re-transmit=
 per share. If we go with Option #1, we&#39;ll definitely have to figure ou=
t a way to mitigate this. (Option #4 is one possibility.)<br></div><div><br=
></div><div>I also agree with Erik and Charlie that Option #2&#39;s downsid=
es can be mitigated by putting an ingestor in front of the helper. Though i=
t&#39;s important to note that this probably cancels out the egress savings=
, since the ingestor (itself a service provider) is proxying report shares =
to the helper. To get the egress savings, the helper would have to be able =
to absorb traffic from clients directly.</div><div><br></div><div>Either wa=
y, Option #2 will require extra protocol logic to deal with synchronization=
 of the report shares. The current spec has each aggregate sub-request from=
 the leader specify the nonce of the report being aggregated. To respond, t=
he helper will first check if it has a report share with that nonce. If not=
 (perhaps the report share hasn&#39;t arrived), it&#39;ll have to treat thi=
s as an error. Where things get tricky is what the leader does next:<br></d=
iv><div>1. It might retry later, but there&#39;s no guarantee that the help=
er&#39;s report share will ever arrive. Eventually it&#39;ll have to give u=
p and the report will get dropped.<br></div><div>2. It might just ignore th=
e report. If the helper&#39;s report share ever shows up, it might end up s=
toring it indefinitely, waiting for the leader to send a request that never=
 arrives.</div><div><br></div><div>This suggests that some data loss may be=
 inevitable in some deployments of Option #2. In my own opinion, this data =
loss is acceptable.<br></div><div><br></div><div>Chris P.<br></div><div><br=
></div><div><br></div><div><br></div><div><br></div><br></div><br><div clas=
s=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 202=
2 at 9:25 AM Phillipp Schoppmann &lt;<a href=3D"mailto:schoppmann@google.co=
m" rel=3D"noreferrer noreferrer noreferrer noreferrer noreferrer" target=3D=
"_blank">schoppmann@google.com</a>&gt; wrote:<br></div><blockquote class=3D=
"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(2=
04,204,204);padding-left:1ex"><div dir=3D"ltr">One more note on the share s=
izes: A lot of overhead (~50%) in our implementation comes from the way I u=
sed protocol buffers for serialization.<div>I think there is some optimizat=
ion potential here, so the cost could be about half that if we use a format=
 that just sends the raw bytes needed.</div><div>For example, with 128 bit =
inputs, the theoretical lower bound of the cost is just 4128 bytes (again, =
excluding the verification part).</div></div><br><div class=3D"gmail_quote"=
><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 6:19 PM Phil=
lipp Schoppmann &lt;<a href=3D"mailto:schoppmann@google.com" rel=3D"norefer=
rer noreferrer noreferrer noreferrer noreferrer" target=3D"_blank">schoppma=
nn@google.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div dir=3D"ltr">Hi everyone,<div><br></div><div>I&#39;m just =
reading through the discussion and catching up.=C2=A0</div><div><a class=3D=
"gmail_plusreply" id=3D"m_-6258409472252836428m_3117631393847772887m_-92351=
6093214407498m_-8313098973848675453m_-2809869134713243354gmail-m_-404417341=
6287307120gmail-m_-5860546362101609079gmail-m_-7075653027769385596gmail-m_-=
7247445031704786523plusReplyChip-0" href=3D"mailto:cpatton@cloudflare.com" =
rel=3D"noreferrer noreferrer noreferrer noreferrer noreferrer" target=3D"_b=
lank">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plusreply" id=3D"m_-62=
58409472252836428m_3117631393847772887m_-923516093214407498m_-8313098973848=
675453m_-2809869134713243354gmail-m_-4044173416287307120gmail-m_-5860546362=
101609079gmail-m_-7075653027769385596gmail-m_-7247445031704786523plusReplyC=
hip-1" href=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer n=
oreferrer noreferrer noreferrer" target=3D"_blank">@Charlie Harrison</a>=C2=
=A0regarding the sizes of IDPF shares, I did some benchmarks:</div><div><ul=
><li>32 bit inputs -&gt; 2127 byte shares</li><li>64 bit inputs -&gt; 4283 =
byte shares</li><li>128 bit inputs -&gt; 8539 byte shares</li></ul></div><d=
iv>These are for a DPF that allows evaluation at every bit of the input (li=
ke Poplar), but currently without the VDAF part. I am in the process of imp=
lementing verification for our IDPF implementation, and can tell you more o=
nce that&#39;s done.</div><div><br></div><div>Best,</div><div>Phillipp</div=
><div><br></div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison &lt;<a hre=
f=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer noreferrer =
noreferrer noreferrer" target=3D"_blank">csharrison@google.com</a>&gt; wrot=
e:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"l=
tr">Thanks Chris,<div>I think I am in agreement with Erik that we could mit=
igate the issues with Option #2 if there is an ingestor client. I think fra=
ming the protocol this way actually allows close imitation=C2=A0of option #=
1 and #2</div><div><br></div><div>- An &quot;option 1&quot; deployment coul=
d=C2=A0just=C2=A0specify one of the aggregators to work as the ingestor eve=
n if it isn&#39;t required by the protocol. They just happen to be the same=
 party.</div><div>- An &quot;option 2&quot; deployment could have the inges=
tor fan out reports to multiple aggregators (or ignore the ingestor complet=
ely)</div><div><br></div><div>Your other points make perfect sense to me!</=
div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_at=
tr">On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a href=3D"mailt=
o:cpatton@cloudflare.com" rel=3D"noreferrer noreferrer noreferrer noreferre=
r noreferrer" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br></=
div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><br>=
=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D=
"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-le=
ft:1ex"><div dir=3D"ltr"><div></div><div>Great! My initial read was that it=
 wouldn&#39;t currently enable it, and I&#39;m happy to work on an issue/PR=
 that can make that more clear and enumerate various deployment=C2=A0scenar=
ios, as you suggest. One point of confusion was the suggestion about limiti=
ng reports from a single IP address, as I&#39;d expect an intermediary serv=
er between the client and leader/aggregators to likely have a fixed IP addr=
ess.</div></div></blockquote><div><br></div><div>Awesome! Let&#39;s start w=
ith opening an issue and folks to start putting their deployment setups the=
re.</div><div><br></div><div>Yes, rate-limiting clients based on IP alone i=
s impossible for aggregators if they&#39;re behind a proxy. It should be no=
ted that this is a heuristic only that an attacker may be able to defeat in=
 many ways. To deal with with Sybil attacks in general, we&#39;ll definitel=
y have to do a bit more anyway.<br></div>=C2=A0</div><div class=3D"gmail_qu=
ote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" rel=3D"noreferrer noreferrer noreferrer nor=
eferrer noreferrer" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer nor=
eferrer noreferrer noreferrer noreferrer noreferrer" target=3D"_blank">http=
s://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div></div>

--00000000000045b9a805d60a4dc4--


From nobody Thu Jan 20 13:46:25 2022
Return-Path: <erik.taubeneck@gmail.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 53B363A2320 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:46:23 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id HRuKAA8Padv4 for <ppm@ietfa.amsl.com>; Thu, 20 Jan 2022 13:46:19 -0800 (PST)
Received: from mail-lf1-x129.google.com (mail-lf1-x129.google.com [IPv6:2a00:1450:4864:20::129]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 16DB63A2326 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:46:18 -0800 (PST)
Received: by mail-lf1-x129.google.com with SMTP id y15so18469697lfa.9 for <ppm@ietf.org>; Thu, 20 Jan 2022 13:46:17 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=VbsiwZYIyp6hE+Lkr002OA3OE+UB0mTLBHBWLyvIzmQ=; b=MUwDRbinb1ZRRRcawCeHxM35JXEZb7gOg6Vo/YS7s+c7Pb2ITXNJGTmwrOhPyjpuQ2 Sxsq9CFaZkd0jnInjZFIUDrkYe9rYII1Er7ZgBx+IBCakbrHAn4f9J92Ndtm3JyfBVHy uwWM39vSku8ukyP8e95+hVDQT6VSLEGUqyNZ6UhejNPW2soxZJrT3GLlm2uQYtj5XyjX 7ecMrDEwU55Nkcpd+gFzFfHjC9tpKoqhA38gGtJq2aiEC20r9bMW2RFoIYJYaU7W3wgb NRFEP0Yz5Sfqxb63AKrK9t51plv7vC7JC/ofMu9UjoFx5Sqp2+6/DBqQj9pbcP79tI8/ n2Rw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=VbsiwZYIyp6hE+Lkr002OA3OE+UB0mTLBHBWLyvIzmQ=; b=doEBlA+9t5/qUKDQF7U3o4NK1mANJ/fGKcS/VtVjd6/HC3sqUiFCkIhAa8nX1+7Ns/ 8eagzJP56DpGsNjtPz9IHpFSTjsh+Miep2kvlHtd+QrcfZot4P9P19x4BBeB9Q7nT+XV 9v3iB8BT2M6Cwe7kbTvNQ+bnbaG0DbRJ6n6hNkXPgUKeDzMz4WtKEF/pc5VfCrd2pwqb nD6uY0GZtfAGkDitTnuT6qT3mnie+b7Ij9heLUgplS1xXyG/dda8klvU/RtApYLMDwZf T/2y8keaYW1pZkvw0rlMkEcME1YTThNzYwl5RmGip2Wc++uiBMLrgyCb4oR2JDc2xYJv bFJQ==
X-Gm-Message-State: AOAM530/mq8gPQRydDKXb9hLoX0jOeB3OxMhko2dhTOcvdduFNzF1lxf J0cg7/fZp7mUXX/5MdB+3xcXehoFfMgz3YF2RDI=
X-Google-Smtp-Source: ABdhPJwgdWmttAIVZADhcBM0LFZJ1io1ZP4snyBmITyvm1KdUkm0WaF7tXxEcmKe1BWwphKgKaC0EW4yDQIibI9as/0=
X-Received: by 2002:a2e:a40b:: with SMTP id p11mr815313ljn.495.1642715174818;  Thu, 20 Jan 2022 13:46:14 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com> <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com>
In-Reply-To: <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com>
From: Erik Taubeneck <erik.taubeneck@gmail.com>
Date: Thu, 20 Jan 2022 13:46:03 -0800
Message-ID: <CANzuiCjDBSaikQJJD39-EZpwSaedue7xq1dZiCJ3ciMruEbdhQ@mail.gmail.com>
To: Watson Ladd <watsonbladd@gmail.com>
Cc: Christopher Patton <cpatton@cloudflare.com>, Charlie Harrison <csharrison@google.com>,  Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>,  ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000007659b705d60a6ef4"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/InAPY5Pfznabwufe3Ikn2FG2ncI>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 20 Jan 2022 21:46:23 -0000

--0000000000007659b705d60a6ef4
Content-Type: text/plain; charset="UTF-8"

>
> Clients to ingestor replication creates many of the same challenges that
> we're trying to solve with the ingestor, in particular partial availability
> and replay if I'm understanding the proposal properly. At this point it's
> not clear why the ingestor and aggregator need to be separate.


I disagree. To me, the point of the ingestor is one of choice and
competition, and not directly one of availability. Consider the example
where a browser vendor implements an API (which is an instance of PPM.)
That browser vendor would need to select some set of aggregators, with
subsets that are trusted to not collude. At that point, if a collector
(independent of the browser vendor) is using that API, it can only select
from those approved subsets to operate as aggregators, giving those
aggregators a fair amount of power. The purpose (that I'm proposing) of the
ingestor (either run by the collector or a 3rd party of their choice) is to
limit the cost borne by the aggregators, so that collectors have as much
choice (and competition exists) in as much of such a protocol as possible.

On Thu, Jan 20, 2022 at 1:36 PM Watson Ladd <watsonbladd@gmail.com> wrote:

>
>
> On Thu, Jan 20, 2022, 4:16 PM Erik Taubeneck <erik.taubeneck@gmail.com>
> wrote:
>
>> To get the egress savings, the helper would have to be able to absorb
>>> traffic from clients directly.
>>>
>>
>> I don't think this is always the case (though it depends on specific
>> cloud providers), but it would require the client to at least send the
>> various shares to different locations. Here is roughly what I envision:
>> helper1 (in cloudA)
>> helper2 (in cloudB)
>>
>> ingestor.proxyA (in cloudA)
>> ingestor.proxyB (in cloudB)
>>
>> If the client is able to send reports to ingestor.proxyA and
>> ingestor.proxyB, then ingestor.proxyA can forward reports to helper1, and
>> ingestor.proxyB can forward reports to helper2.  There should be ways in
>> many cloud providers for that forwarding to not incur egress fees (for
>> example, AWS allows that to happen with S3 so long as both accounts are
>> within the same region.)
>>
>
> Clients to ingestor replication creates many of the same challenges that
> we're trying to solve with the ingestor, in particular partial availability
> and replay if I'm understanding the proposal properly. At this point it's
> not clear why the ingestor and aggregator need to be separate.
>
>>
>> I'm not sure if this is compatible with what you envision for Option #2,
>> but it does seem desirable for an ingestor to at least have this option for
>> cost reduction.
>>
>
>> Erik
>>
>> On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton <
>> cpatton@cloudflare.com> wrote:
>>
>>> Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
>>> encoding for the input shares (including the IDPF key shares), which we can
>>> reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
>>> of data to re-transmit per share. If we go with Option #1, we'll definitely
>>> have to figure out a way to mitigate this. (Option #4 is one possibility.)
>>>
>>> I also agree with Erik and Charlie that Option #2's downsides can be
>>> mitigated by putting an ingestor in front of the helper. Though it's
>>> important to note that this probably cancels out the egress savings, since
>>> the ingestor (itself a service provider) is proxying report shares to the
>>> helper. To get the egress savings, the helper would have to be able to
>>> absorb traffic from clients directly.
>>>
>>> Either way, Option #2 will require extra protocol logic to deal with
>>> synchronization of the report shares. The current spec has each aggregate
>>> sub-request from the leader specify the nonce of the report being
>>> aggregated. To respond, the helper will first check if it has a report
>>> share with that nonce. If not (perhaps the report share hasn't arrived),
>>> it'll have to treat this as an error. Where things get tricky is what the
>>> leader does next:
>>> 1. It might retry later, but there's no guarantee that the helper's
>>> report share will ever arrive. Eventually it'll have to give up and the
>>> report will get dropped.
>>> 2. It might just ignore the report. If the helper's report share ever
>>> shows up, it might end up storing it indefinitely, waiting for the leader
>>> to send a request that never arrives.
>>>
>>> This suggests that some data loss may be inevitable in some deployments
>>> of Option #2. In my own opinion, this data loss is acceptable.
>>>
>>> Chris P.
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <
>>> schoppmann@google.com> wrote:
>>>
>>>> One more note on the share sizes: A lot of overhead (~50%) in our
>>>> implementation comes from the way I used protocol buffers for serialization.
>>>> I think there is some optimization potential here, so the cost could be
>>>> about half that if we use a format that just sends the raw bytes needed.
>>>> For example, with 128 bit inputs, the theoretical lower bound of the
>>>> cost is just 4128 bytes (again, excluding the verification part).
>>>>
>>>> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <
>>>> schoppmann@google.com> wrote:
>>>>
>>>>> Hi everyone,
>>>>>
>>>>> I'm just reading through the discussion and catching up.
>>>>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>>>>> <csharrison@google.com> regarding the sizes of IDPF shares, I did
>>>>> some benchmarks:
>>>>>
>>>>>    - 32 bit inputs -> 2127 byte shares
>>>>>    - 64 bit inputs -> 4283 byte shares
>>>>>    - 128 bit inputs -> 8539 byte shares
>>>>>
>>>>> These are for a DPF that allows evaluation at every bit of the input
>>>>> (like Poplar), but currently without the VDAF part. I am in the process of
>>>>> implementing verification for our IDPF implementation, and can tell you
>>>>> more once that's done.
>>>>>
>>>>> Best,
>>>>> Phillipp
>>>>>
>>>>>
>>>>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <
>>>>> csharrison@google.com> wrote:
>>>>>
>>>>>> Thanks Chris,
>>>>>> I think I am in agreement with Erik that we could mitigate the issues
>>>>>> with Option #2 if there is an ingestor client. I think framing the protocol
>>>>>> this way actually allows close imitation of option #1 and #2
>>>>>>
>>>>>> - An "option 1" deployment could just specify one of the aggregators
>>>>>> to work as the ingestor even if it isn't required by the protocol. They
>>>>>> just happen to be the same party.
>>>>>> - An "option 2" deployment could have the ingestor fan out reports to
>>>>>> multiple aggregators (or ignore the ingestor completely)
>>>>>>
>>>>>> Your other points make perfect sense to me!
>>>>>>
>>>>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>>>>> cpatton@cloudflare.com> wrote:
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> Great! My initial read was that it wouldn't currently enable it,
>>>>>>>> and I'm happy to work on an issue/PR that can make that more clear and
>>>>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>>>>> address, as I'd expect an intermediary server between the client and
>>>>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>>>>
>>>>>>>
>>>>>>> Awesome! Let's start with opening an issue and folks to start
>>>>>>> putting their deployment setups there.
>>>>>>>
>>>>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>>>>> anyway.
>>>>>>>
>>>>>>>
>>>>>>> Chris P.
>>>>>>>
>>>>>> --
>> Ppm mailing list
>> Ppm@ietf.org
>> https://www.ietf.org/mailman/listinfo/ppm
>>
>

--0000000000007659b705d60a6ef4
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Clients =
to ingestor replication creates many of the same challenges that we&#39;re =
trying to solve with the ingestor, in particular partial availability and r=
eplay if I&#39;m understanding the proposal properly. At this point it&#39;=
s not clear why the ingestor and aggregator need to be separate.</blockquot=
e><div><br></div><div>I disagree. To me, the point of the ingestor is one o=
f choice and competition, and not directly one of availability. Consider th=
e example where a browser vendor implements an API (which is an instance of=
 PPM.) That browser vendor would need to select some set of aggregators, wi=
th subsets that are trusted to not collude. At that point, if a collector (=
independent of the browser vendor) is using that API, it can only select fr=
om those approved subsets to operate as aggregators, giving those aggregato=
rs a fair amount of power. The purpose (that I&#39;m proposing) of the inge=
stor (either run by the collector or a 3rd party of their choice) is to lim=
it the cost borne by the aggregators, so that collectors have as much choic=
e (and competition exists) in as much of such a protocol as possible.=C2=A0=
</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_=
attr">On Thu, Jan 20, 2022 at 1:36 PM Watson Ladd &lt;<a href=3D"mailto:wat=
sonbladd@gmail.com">watsonbladd@gmail.com</a>&gt; wrote:<br></div><blockquo=
te class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px =
solid rgb(204,204,204);padding-left:1ex"><div dir=3D"auto"><div><br><br><di=
v class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 2=
0, 2022, 4:16 PM Erik Taubeneck &lt;<a href=3D"mailto:erik.taubeneck@gmail.=
com" rel=3D"noreferrer noreferrer noreferrer noreferrer" target=3D"_blank">=
erik.taubeneck@gmail.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex"><div dir=3D"ltr"><blockquote class=3D"gmail_quote" =
style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);pa=
dding-left:1ex">To get the egress savings, the helper would have to be able=
 to absorb traffic from clients directly.<br></blockquote><div><br></div><d=
iv>I don&#39;t think this is always the case (though=C2=A0it depends on spe=
cific cloud providers), but it would require the client to at least send th=
e various=C2=A0shares to different locations. Here is roughly what I envisi=
on:</div><div>helper1 (in cloudA)</div><div>helper2 (in cloudB)</div><div><=
br></div><div>ingestor.proxyA (in cloudA)</div><div>ingestor.proxyB (in clo=
udB)</div><div><br></div><div>If the client is able to send reports to inge=
stor.proxyA and ingestor.proxyB, then ingestor.proxyA can forward reports t=
o helper1, and ingestor.proxyB can forward reports to helper2.=C2=A0 There =
should be ways in many cloud providers for that forwarding to not incur egr=
ess fees (for example, AWS allows that to happen with S3 so long as both ac=
counts are within the same region.)</div></div></blockquote></div></div><di=
v dir=3D"auto"><br></div><div dir=3D"auto">Clients to ingestor replication =
creates many of the same challenges that we&#39;re trying to solve with the=
 ingestor, in particular partial availability and replay if I&#39;m underst=
anding the proposal properly. At this point it&#39;s not clear why the inge=
stor and aggregator need to be separate.</div><div dir=3D"auto"><div class=
=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=
=3D"ltr"><div><br></div><div>I&#39;m not sure if this is compatible with wh=
at you envision for Option #2, but it does seem desirable for an ingestor t=
o at least have this option for cost reduction.</div></div></blockquote></d=
iv></div><div dir=3D"auto"><div class=3D"gmail_quote"><blockquote class=3D"=
gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(20=
4,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br></div><div>Erik</div=
></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr"=
>On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton &lt;<a href=3D"mailto:=
cpatton@cloudflare.com" rel=3D"noreferrer noreferrer noreferrer noreferrer =
noreferrer" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br></di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>T=
hanks for the numbers, Philllipp! The VDAF spec will specify a binary encod=
ing for the input shares (including the IDPF key shares), which we can reas=
onably expect to be much closer to the lower bound. Still, 4KB is a lot of =
data to re-transmit per share. If we go with Option #1, we&#39;ll definitel=
y have to figure out a way to mitigate this. (Option #4 is one possibility.=
)<br></div><div><br></div><div>I also agree with Erik and Charlie that Opti=
on #2&#39;s downsides can be mitigated by putting an ingestor in front of t=
he helper. Though it&#39;s important to note that this probably cancels out=
 the egress savings, since the ingestor (itself a service provider) is prox=
ying report shares to the helper. To get the egress savings, the helper wou=
ld have to be able to absorb traffic from clients directly.</div><div><br><=
/div><div>Either way, Option #2 will require extra protocol logic to deal w=
ith synchronization of the report shares. The current spec has each aggrega=
te sub-request from the leader specify the nonce of the report being aggreg=
ated. To respond, the helper will first check if it has a report share with=
 that nonce. If not (perhaps the report share hasn&#39;t arrived), it&#39;l=
l have to treat this as an error. Where things get tricky is what the leade=
r does next:<br></div><div>1. It might retry later, but there&#39;s no guar=
antee that the helper&#39;s report share will ever arrive. Eventually it&#3=
9;ll have to give up and the report will get dropped.<br></div><div>2. It m=
ight just ignore the report. If the helper&#39;s report share ever shows up=
, it might end up storing it indefinitely, waiting for the leader to send a=
 request that never arrives.</div><div><br></div><div>This suggests that so=
me data loss may be inevitable in some deployments of Option #2. In my own =
opinion, this data loss is acceptable.<br></div><div><br></div><div>Chris P=
.<br></div><div><br></div><div><br></div><div><br></div><div><br></div><br>=
</div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">=
On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann &lt;<a href=3D"mailto:s=
choppmann@google.com" rel=3D"noreferrer noreferrer noreferrer noreferrer no=
referrer" target=3D"_blank">schoppmann@google.com</a>&gt; wrote:<br></div><=
blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-l=
eft:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">One more =
note on the share sizes: A lot of overhead (~50%) in our implementation com=
es from the way I used protocol buffers for serialization.<div>I think ther=
e is some optimization potential here, so the cost could be about half that=
 if we use a format that just sends the raw bytes needed.</div><div>For exa=
mple, with 128 bit inputs, the theoretical lower bound of the cost is just =
4128 bytes (again, excluding the verification part).</div></div><br><div cl=
ass=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2=
022 at 6:19 PM Phillipp Schoppmann &lt;<a href=3D"mailto:schoppmann@google.=
com" rel=3D"noreferrer noreferrer noreferrer noreferrer noreferrer" target=
=3D"_blank">schoppmann@google.com</a>&gt; wrote:<br></div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr">Hi everyone,<div><br></di=
v><div>I&#39;m just reading through the discussion and catching up.=C2=A0</=
div><div><a class=3D"gmail_plusreply" id=3D"gmail-m_-8191592490242430118m_-=
6258409472252836428m_3117631393847772887m_-923516093214407498m_-83130989738=
48675453m_-2809869134713243354gmail-m_-4044173416287307120gmail-m_-58605463=
62101609079gmail-m_-7075653027769385596gmail-m_-7247445031704786523plusRepl=
yChip-0" href=3D"mailto:cpatton@cloudflare.com" rel=3D"noreferrer noreferre=
r noreferrer noreferrer noreferrer" target=3D"_blank">@Christopher Patton</=
a>=C2=A0<a class=3D"gmail_plusreply" id=3D"gmail-m_-8191592490242430118m_-6=
258409472252836428m_3117631393847772887m_-923516093214407498m_-831309897384=
8675453m_-2809869134713243354gmail-m_-4044173416287307120gmail-m_-586054636=
2101609079gmail-m_-7075653027769385596gmail-m_-7247445031704786523plusReply=
Chip-1" href=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer =
noreferrer noreferrer noreferrer" target=3D"_blank">@Charlie Harrison</a>=
=C2=A0regarding the sizes of IDPF shares, I did some benchmarks:</div><div>=
<ul><li>32 bit inputs -&gt; 2127 byte shares</li><li>64 bit inputs -&gt; 42=
83 byte shares</li><li>128 bit inputs -&gt; 8539 byte shares</li></ul></div=
><div>These are for a DPF that allows evaluation at every bit of the input =
(like Poplar), but currently without the VDAF part. I am in the process of =
implementing verification for our IDPF implementation, and can tell you mor=
e once that&#39;s done.</div><div><br></div><div>Best,</div><div>Phillipp</=
div><div><br></div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" cl=
ass=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison &lt;<a =
href=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer noreferr=
er noreferrer noreferrer" target=3D"_blank">csharrison@google.com</a>&gt; w=
rote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0p=
x 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=
=3D"ltr">Thanks Chris,<div>I think I am in agreement with Erik that we coul=
d mitigate the issues with Option #2 if there is an ingestor client. I thin=
k framing the protocol this way actually allows close imitation=C2=A0of opt=
ion #1 and #2</div><div><br></div><div>- An &quot;option 1&quot; deployment=
 could=C2=A0just=C2=A0specify one of the aggregators to work as the ingesto=
r even if it isn&#39;t required by the protocol. They just happen to be the=
 same party.</div><div>- An &quot;option 2&quot; deployment could have the =
ingestor fan out reports to multiple aggregators (or ignore the ingestor co=
mpletely)</div><div><br></div><div>Your other points make perfect sense to =
me!</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a href=3D"=
mailto:cpatton@cloudflare.com" rel=3D"noreferrer noreferrer noreferrer nore=
ferrer noreferrer" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<=
br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8e=
x;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"=
><br>=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div dir=3D"ltr"><div></div><div>Great! My initial read was th=
at it wouldn&#39;t currently enable it, and I&#39;m happy to work on an iss=
ue/PR that can make that more clear and enumerate various deployment=C2=A0s=
cenarios, as you suggest. One point of confusion was the suggestion about l=
imiting reports from a single IP address, as I&#39;d expect an intermediary=
 server between the client and leader/aggregators to likely have a fixed IP=
 address.</div></div></blockquote><div><br></div><div>Awesome! Let&#39;s st=
art with opening an issue and folks to start putting their deployment setup=
s there.</div><div><br></div><div>Yes, rate-limiting clients based on IP al=
one is impossible for aggregators if they&#39;re behind a proxy. It should =
be noted that this is a heuristic only that an attacker may be able to defe=
at in many ways. To deal with with Sybil attacks in general, we&#39;ll defi=
nitely have to do a bit more anyway.<br></div>=C2=A0</div><div class=3D"gma=
il_quote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" rel=3D"noreferrer noreferrer noreferrer nor=
eferrer noreferrer" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer nor=
eferrer noreferrer noreferrer noreferrer noreferrer" target=3D"_blank">http=
s://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div></div>
</blockquote></div>

--0000000000007659b705d60a6ef4--


From nobody Fri Jan 21 17:40:36 2022
Return-Path: <timg@letsencrypt.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id A265A3A1B75 for <ppm@ietfa.amsl.com>; Fri, 21 Jan 2022 17:40:35 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.674
X-Spam-Level: 
X-Spam-Status: No, score=-2.674 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.576, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=letsencrypt.org
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id IOsMhMh9Q0ra for <ppm@ietfa.amsl.com>; Fri, 21 Jan 2022 17:40:30 -0800 (PST)
Received: from mail-qt1-x82e.google.com (mail-qt1-x82e.google.com [IPv6:2607:f8b0:4864:20::82e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id B40ED3A1B73 for <ppm@ietf.org>; Fri, 21 Jan 2022 17:40:30 -0800 (PST)
Received: by mail-qt1-x82e.google.com with SMTP id m9so8277678qtq.10 for <ppm@ietf.org>; Fri, 21 Jan 2022 17:40:30 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=letsencrypt.org; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=Qb2gS/M8FJcY/ozIrnzPjxBwkjM8wFlr/PlEequAGrw=; b=cUmryFpH+l38vBhyyItZMdz9lOuHdJCXBZ9VuTXDDmivid2t+9YG7f7UDjkIvwdJR6 5t8X04diqK+mYOSNVIcL+nmxx/Q4b8FzGbvm9FOQziQeayektiw51yk3E6kv7mMyDWzg Zw2og8K40Bnb922Kglhi1751dMuPlsipzkfc0=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=Qb2gS/M8FJcY/ozIrnzPjxBwkjM8wFlr/PlEequAGrw=; b=5z3cWa/yewsnh3KwrN/HOtdPu3L40C0x7CQD3IB0aEjuuKgbC6jBHcgus83AhMGESX 1/BtyhT5RyTPZsBDBfIAgBQoqr3UMYeW0L9gZVKyhTMk40op364LIFsDYB83YOjxd3JE vaQMcXm8K93dxwx+jenL02O+eTMAKSKJCDzrR/gv1npqR7z+GzmKzcLj0cVzxBf+ZGPo 2FcLxvVdxQE6/V/18/GtMsT04AZLEhHoFidTGTDJVk1AIXW2lAYRsOi2gwbkdseqYHcd gSy98I4xCWqgsEyul+0eta8akgx7xcf1cvXV7oScCcD8Jb9SdJeuYOdun81tel2Ysw+9 oLvA==
X-Gm-Message-State: AOAM530FsSi+1Lfkf8STvr4DDunmfv2LElYbMRux8VqzgZtSwAPSxOu9 0EVwB9BRTDD2aK4F0r0eHBW+tZT6kBLMalpelrlDyg==
X-Google-Smtp-Source: ABdhPJzYInzx1MaRQYvfMxcz2mDYcSU3lLnH/rbZj/kaxbqEJXg6aZ7HxkKgGy6B3aL7bKd2Zs72sPGSZT7LHaxWGAc=
X-Received: by 2002:ac8:7202:: with SMTP id a2mr5604254qtp.268.1642815628906;  Fri, 21 Jan 2022 17:40:28 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com> <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com> <CANzuiCjDBSaikQJJD39-EZpwSaedue7xq1dZiCJ3ciMruEbdhQ@mail.gmail.com>
In-Reply-To: <CANzuiCjDBSaikQJJD39-EZpwSaedue7xq1dZiCJ3ciMruEbdhQ@mail.gmail.com>
From: Tim Geoghegan <timg@letsencrypt.org>
Date: Fri, 21 Jan 2022 17:40:17 -0800
Message-ID: <CABN231rmmXmaSDxhXwX2TG+DVqDjHVWkoVVxUZsC2oQOv8X20g@mail.gmail.com>
To: Erik Taubeneck <erik.taubeneck@gmail.com>
Cc: Watson Ladd <watsonbladd@gmail.com>, Christopher Patton <cpatton@cloudflare.com>, ppm@ietf.org,  Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>,  Charlie Harrison <csharrison@google.com>
Content-Type: multipart/alternative; boundary="000000000000fe25ff05d621d1be"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/OSfhcossIrEPqPhIEW5CkVcJhik>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sat, 22 Jan 2022 01:40:36 -0000

--000000000000fe25ff05d621d1be
Content-Type: text/plain; charset="UTF-8"

I want to chime in on the question of single upload (what we spell out in
the spec today) vs. split upload (i.e., client is responsible for
separately uploading report shares to each aggregator, or a proxy for the
aggregator). In my view, there are two competing concerns to weigh:
simplicity of the client vs. the egress cost problem.

Having clients upload both report shares to the leader in one request makes
things nice and transactional: clients perform one upload and if they get
HTTP 200 OK, they are assured that the report has been stored by the leader
and so the client can discard it and move on. Also the leader can guarantee
that both report shares are available to be aggregated, because it holds
both of them. The downside is that the leader incurs egress costs when it
retransmits shares to the helper.

If the client has to upload each share to a different aggregator, though,
the transaction is messier: what happens if the upload to the leader
succeeds, but the one to the helper fails (or vice versa)? That's more
state for the client to juggle (and I think simplicity in the client is a
virtue we should aim for), but it also means the aggregators have to deal
with the possibility of the helper and leader not having the same set of
shares.

A few of us were batting this around earlier and concluded that the problem
of aggregators having disjoint views of what shares exist occurs no matter
what: even in the single upload model, we would expect that aggregators
will implement some kind of retention policy for submitted reports to stop
their storage costs from growing indefinitely. So maybe a helper discards
unaggregated shares after 14 days but its leader does so after 21. Even if
they use the same retention policy, we can't guarantee that they will
discard shares at the same moment. If a collector takes long enough to
issue a `CollectReq` (in some VDAFs like `poplar1`, the aggregate phase
cannot begin until the collector supplies the aggregation parameter, i.e.,
the candidate prefixes), then we are in the same error case that is
introduced by partial success of a split upload. That means that the PPM
protocol has to handle this error case no matter what (and we believe it
already does).

Split uploads provides a natural solution to the egress cost problem, as
explained upthread. But Erik T. raises some good points about asymmetries
in availability between aggregators, so we also want a solution that
enables one aggregator, run by an organization that doesn't flinch at
providing ninety nine nines of availability and is willing to accept egress
costs, to field all upload requests from clients. I'm against option #3
(allowing deployments to use either single or split uploads), because I
would prefer that the protocol be as simple as possible. In a split upload
model, the task parameters would have to include two different `upload`
endpoints to which clients send shares. Deployments could simply have both
upload endpoints be hosted by the same incredibly available server to spare
the other aggregator the need to scale up or attain the required uptime.

Split uploads do incur some client complexity I'd rather avoid and they do
increase the chance of upload failures (because there are two requests than
can fail instead of just one), but I think that model supports the greatest
diversity of deployments with the least complexity in the protocol

Tim Geoghegan
ISRG

On Thu, Jan 20, 2022 at 1:46 PM Erik Taubeneck <erik.taubeneck@gmail.com>
wrote:

> Clients to ingestor replication creates many of the same challenges that
>> we're trying to solve with the ingestor, in particular partial availability
>> and replay if I'm understanding the proposal properly. At this point it's
>> not clear why the ingestor and aggregator need to be separate.
>
>
> I disagree. To me, the point of the ingestor is one of choice and
> competition, and not directly one of availability. Consider the example
> where a browser vendor implements an API (which is an instance of PPM.)
> That browser vendor would need to select some set of aggregators, with
> subsets that are trusted to not collude. At that point, if a collector
> (independent of the browser vendor) is using that API, it can only select
> from those approved subsets to operate as aggregators, giving those
> aggregators a fair amount of power. The purpose (that I'm proposing) of the
> ingestor (either run by the collector or a 3rd party of their choice) is to
> limit the cost borne by the aggregators, so that collectors have as much
> choice (and competition exists) in as much of such a protocol as possible.
>
> On Thu, Jan 20, 2022 at 1:36 PM Watson Ladd <watsonbladd@gmail.com> wrote:
>
>>
>>
>> On Thu, Jan 20, 2022, 4:16 PM Erik Taubeneck <erik.taubeneck@gmail.com>
>> wrote:
>>
>>> To get the egress savings, the helper would have to be able to absorb
>>>> traffic from clients directly.
>>>>
>>>
>>> I don't think this is always the case (though it depends on specific
>>> cloud providers), but it would require the client to at least send the
>>> various shares to different locations. Here is roughly what I envision:
>>> helper1 (in cloudA)
>>> helper2 (in cloudB)
>>>
>>> ingestor.proxyA (in cloudA)
>>> ingestor.proxyB (in cloudB)
>>>
>>> If the client is able to send reports to ingestor.proxyA and
>>> ingestor.proxyB, then ingestor.proxyA can forward reports to helper1, and
>>> ingestor.proxyB can forward reports to helper2.  There should be ways in
>>> many cloud providers for that forwarding to not incur egress fees (for
>>> example, AWS allows that to happen with S3 so long as both accounts are
>>> within the same region.)
>>>
>>
>> Clients to ingestor replication creates many of the same challenges that
>> we're trying to solve with the ingestor, in particular partial availability
>> and replay if I'm understanding the proposal properly. At this point it's
>> not clear why the ingestor and aggregator need to be separate.
>>
>>>
>>> I'm not sure if this is compatible with what you envision for Option #2,
>>> but it does seem desirable for an ingestor to at least have this option for
>>> cost reduction.
>>>
>>
>>> Erik
>>>
>>> On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton <
>>> cpatton@cloudflare.com> wrote:
>>>
>>>> Thanks for the numbers, Philllipp! The VDAF spec will specify a binary
>>>> encoding for the input shares (including the IDPF key shares), which we can
>>>> reasonably expect to be much closer to the lower bound. Still, 4KB is a lot
>>>> of data to re-transmit per share. If we go with Option #1, we'll definitely
>>>> have to figure out a way to mitigate this. (Option #4 is one possibility.)
>>>>
>>>> I also agree with Erik and Charlie that Option #2's downsides can be
>>>> mitigated by putting an ingestor in front of the helper. Though it's
>>>> important to note that this probably cancels out the egress savings, since
>>>> the ingestor (itself a service provider) is proxying report shares to the
>>>> helper. To get the egress savings, the helper would have to be able to
>>>> absorb traffic from clients directly.
>>>>
>>>> Either way, Option #2 will require extra protocol logic to deal with
>>>> synchronization of the report shares. The current spec has each aggregate
>>>> sub-request from the leader specify the nonce of the report being
>>>> aggregated. To respond, the helper will first check if it has a report
>>>> share with that nonce. If not (perhaps the report share hasn't arrived),
>>>> it'll have to treat this as an error. Where things get tricky is what the
>>>> leader does next:
>>>> 1. It might retry later, but there's no guarantee that the helper's
>>>> report share will ever arrive. Eventually it'll have to give up and the
>>>> report will get dropped.
>>>> 2. It might just ignore the report. If the helper's report share ever
>>>> shows up, it might end up storing it indefinitely, waiting for the leader
>>>> to send a request that never arrives.
>>>>
>>>> This suggests that some data loss may be inevitable in some deployments
>>>> of Option #2. In my own opinion, this data loss is acceptable.
>>>>
>>>> Chris P.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann <
>>>> schoppmann@google.com> wrote:
>>>>
>>>>> One more note on the share sizes: A lot of overhead (~50%) in our
>>>>> implementation comes from the way I used protocol buffers for serialization.
>>>>> I think there is some optimization potential here, so the cost could
>>>>> be about half that if we use a format that just sends the raw bytes needed.
>>>>> For example, with 128 bit inputs, the theoretical lower bound of the
>>>>> cost is just 4128 bytes (again, excluding the verification part).
>>>>>
>>>>> On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann <
>>>>> schoppmann@google.com> wrote:
>>>>>
>>>>>> Hi everyone,
>>>>>>
>>>>>> I'm just reading through the discussion and catching up.
>>>>>> @Christopher Patton <cpatton@cloudflare.com> @Charlie Harrison
>>>>>> <csharrison@google.com> regarding the sizes of IDPF shares, I did
>>>>>> some benchmarks:
>>>>>>
>>>>>>    - 32 bit inputs -> 2127 byte shares
>>>>>>    - 64 bit inputs -> 4283 byte shares
>>>>>>    - 128 bit inputs -> 8539 byte shares
>>>>>>
>>>>>> These are for a DPF that allows evaluation at every bit of the input
>>>>>> (like Poplar), but currently without the VDAF part. I am in the process of
>>>>>> implementing verification for our IDPF implementation, and can tell you
>>>>>> more once that's done.
>>>>>>
>>>>>> Best,
>>>>>> Phillipp
>>>>>>
>>>>>>
>>>>>> On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison <
>>>>>> csharrison@google.com> wrote:
>>>>>>
>>>>>>> Thanks Chris,
>>>>>>> I think I am in agreement with Erik that we could mitigate the
>>>>>>> issues with Option #2 if there is an ingestor client. I think framing the
>>>>>>> protocol this way actually allows close imitation of option #1 and #2
>>>>>>>
>>>>>>> - An "option 1" deployment could just specify one of the aggregators
>>>>>>> to work as the ingestor even if it isn't required by the protocol. They
>>>>>>> just happen to be the same party.
>>>>>>> - An "option 2" deployment could have the ingestor fan out reports
>>>>>>> to multiple aggregators (or ignore the ingestor completely)
>>>>>>>
>>>>>>> Your other points make perfect sense to me!
>>>>>>>
>>>>>>> On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton <
>>>>>>> cpatton@cloudflare.com> wrote:
>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> Great! My initial read was that it wouldn't currently enable it,
>>>>>>>>> and I'm happy to work on an issue/PR that can make that more clear and
>>>>>>>>> enumerate various deployment scenarios, as you suggest. One point of
>>>>>>>>> confusion was the suggestion about limiting reports from a single IP
>>>>>>>>> address, as I'd expect an intermediary server between the client and
>>>>>>>>> leader/aggregators to likely have a fixed IP address.
>>>>>>>>>
>>>>>>>>
>>>>>>>> Awesome! Let's start with opening an issue and folks to start
>>>>>>>> putting their deployment setups there.
>>>>>>>>
>>>>>>>> Yes, rate-limiting clients based on IP alone is impossible for
>>>>>>>> aggregators if they're behind a proxy. It should be noted that this is a
>>>>>>>> heuristic only that an attacker may be able to defeat in many ways. To deal
>>>>>>>> with with Sybil attacks in general, we'll definitely have to do a bit more
>>>>>>>> anyway.
>>>>>>>>
>>>>>>>>
>>>>>>>> Chris P.
>>>>>>>>
>>>>>>> --
>>> Ppm mailing list
>>> Ppm@ietf.org
>>> https://www.ietf.org/mailman/listinfo/ppm
>>>
>> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--000000000000fe25ff05d621d1be
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>I want to chime in on the question of single upload (=
what we spell out in the spec today) vs. split upload (i.e., client is resp=
onsible for separately uploading report shares to each aggregator, or a pro=
xy for the aggregator). In my view, there are two competing concerns to wei=
gh: simplicity of the client vs. the egress cost problem.</div><div><br></d=
iv><div>Having clients upload both report shares to the leader in one reque=
st makes things nice and transactional: clients perform one upload and if t=
hey get HTTP 200 OK, they are assured that the report has been stored by th=
e leader and so the client can discard it and move on. Also the leader can =
guarantee that both report shares are available to be aggregated, because i=
t holds both of them. The downside is that the leader incurs egress costs w=
hen it retransmits shares to the helper.</div><div><br></div><div>If the cl=
ient has to upload each share to a different aggregator, though, the transa=
ction is messier: what happens if the upload to the leader succeeds, but th=
e one to the helper fails (or vice versa)? That&#39;s more state for the cl=
ient to juggle (and I think simplicity in the client is a virtue we should =
aim for), but it also means the aggregators have to deal with the possibili=
ty of the helper and leader not having the same set of shares.</div><div><b=
r></div><div>A few of us were batting this around earlier and concluded tha=
t the problem of aggregators having disjoint views of what shares exist occ=
urs no matter what: even in the single upload model, we would expect that a=
ggregators will implement some kind of retention policy for submitted repor=
ts to stop their storage costs from growing indefinitely. So maybe a helper=
 discards unaggregated shares after 14 days but its leader does so after 21=
. Even if they use the same retention policy, we can&#39;t guarantee that t=
hey will discard shares at the same moment. If a collector takes long enoug=
h to issue a `CollectReq` (in some VDAFs like `poplar1`, the aggregate phas=
e cannot begin until the collector supplies the aggregation parameter, i.e.=
, the candidate prefixes), then we are in the same error case that is intro=
duced by partial success of a split upload. That means that the PPM protoco=
l has to handle this error case no matter what (and we believe it already d=
oes).<br></div><div><br></div><div>Split uploads provides a natural solutio=
n to the egress cost problem, as explained upthread. But Erik T. raises som=
e good points about asymmetries in availability between aggregators, so we =
also want a solution that enables one aggregator, run by an organization th=
at doesn&#39;t flinch at providing ninety nine nines of availability and is=
 willing to accept egress costs, to field all upload requests from clients.=
 I&#39;m against option #3 (allowing deployments to use either single or sp=
lit uploads), because I would prefer that the protocol be as simple as poss=
ible. In a split upload model, the task parameters would have to include tw=
o different `upload` endpoints to which clients send shares. Deployments co=
uld simply have both upload endpoints be hosted by the same incredibly avai=
lable server to spare the other aggregator the need to scale up or attain t=
he required uptime.</div><div><br></div><div>Split uploads do incur some cl=
ient complexity I&#39;d rather avoid and they do increase the chance of upl=
oad failures (because there are two requests than can fail instead of just =
one), but I think that model supports the greatest diversity of deployments=
 with the least complexity in the protocol</div><div><br></div><div>Tim Geo=
ghegan</div><div>ISRG<br></div></div><br><div class=3D"gmail_quote"><div di=
r=3D"ltr" class=3D"gmail_attr">On Thu, Jan 20, 2022 at 1:46 PM Erik Taubene=
ck &lt;<a href=3D"mailto:erik.taubeneck@gmail.com">erik.taubeneck@gmail.com=
</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:=
0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">=
<div dir=3D"ltr"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Clients =
to ingestor replication creates many of the same challenges that we&#39;re =
trying to solve with the ingestor, in particular partial availability and r=
eplay if I&#39;m understanding the proposal properly. At this point it&#39;=
s not clear why the ingestor and aggregator need to be separate.</blockquot=
e><div><br></div><div>I disagree. To me, the point of the ingestor is one o=
f choice and competition, and not directly one of availability. Consider th=
e example where a browser vendor implements an API (which is an instance of=
 PPM.) That browser vendor would need to select some set of aggregators, wi=
th subsets that are trusted to not collude. At that point, if a collector (=
independent of the browser vendor) is using that API, it can only select fr=
om those approved subsets to operate as aggregators, giving those aggregato=
rs a fair amount of power. The purpose (that I&#39;m proposing) of the inge=
stor (either run by the collector or a 3rd party of their choice) is to lim=
it the cost borne by the aggregators, so that collectors have as much choic=
e (and competition exists) in as much of such a protocol as possible.=C2=A0=
</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_=
attr">On Thu, Jan 20, 2022 at 1:36 PM Watson Ladd &lt;<a href=3D"mailto:wat=
sonbladd@gmail.com" target=3D"_blank">watsonbladd@gmail.com</a>&gt; wrote:<=
br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8e=
x;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"auto=
"><div><br><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_a=
ttr">On Thu, Jan 20, 2022, 4:16 PM Erik Taubeneck &lt;<a href=3D"mailto:eri=
k.taubeneck@gmail.com" rel=3D"noreferrer noreferrer noreferrer noreferrer" =
target=3D"_blank">erik.taubeneck@gmail.com</a>&gt; wrote:<br></div><blockqu=
ote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px=
 solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex">To get the egress savings, the helper wou=
ld have to be able to absorb traffic from clients directly.<br></blockquote=
><div><br></div><div>I don&#39;t think this is always the case (though=C2=
=A0it depends on specific cloud providers), but it would require the client=
 to at least send the various=C2=A0shares to different locations. Here is r=
oughly what I envision:</div><div>helper1 (in cloudA)</div><div>helper2 (in=
 cloudB)</div><div><br></div><div>ingestor.proxyA (in cloudA)</div><div>ing=
estor.proxyB (in cloudB)</div><div><br></div><div>If the client is able to =
send reports to ingestor.proxyA and ingestor.proxyB, then ingestor.proxyA c=
an forward reports to helper1, and ingestor.proxyB can forward reports to h=
elper2.=C2=A0 There should be ways in many cloud providers for that forward=
ing to not incur egress fees (for example, AWS allows that to happen with S=
3 so long as both accounts are within the same region.)</div></div></blockq=
uote></div></div><div dir=3D"auto"><br></div><div dir=3D"auto">Clients to i=
ngestor replication creates many of the same challenges that we&#39;re tryi=
ng to solve with the ingestor, in particular partial availability and repla=
y if I&#39;m understanding the proposal properly. At this point it&#39;s no=
t clear why the ingestor and aggregator need to be separate.</div><div dir=
=3D"auto"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" styl=
e=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddin=
g-left:1ex"><div dir=3D"ltr"><div><br></div><div>I&#39;m not sure if this i=
s compatible with what you envision for Option #2, but it does seem desirab=
le for an ingestor to at least have this option for cost reduction.</div></=
div></blockquote></div></div><div dir=3D"auto"><div class=3D"gmail_quote"><=
blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-l=
eft:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br>=
</div><div>Erik</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" =
class=3D"gmail_attr">On Thu, Jan 20, 2022 at 11:18 AM Christopher Patton &l=
t;<a href=3D"mailto:cpatton@cloudflare.com" rel=3D"noreferrer noreferrer no=
referrer noreferrer noreferrer" target=3D"_blank">cpatton@cloudflare.com</a=
>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px=
 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><di=
v dir=3D"ltr"><div>Thanks for the numbers, Philllipp! The VDAF spec will sp=
ecify a binary encoding for the input shares (including the IDPF key shares=
), which we can reasonably expect to be much closer to the lower bound. Sti=
ll, 4KB is a lot of data to re-transmit per share. If we go with Option #1,=
 we&#39;ll definitely have to figure out a way to mitigate this. (Option #4=
 is one possibility.)<br></div><div><br></div><div>I also agree with Erik a=
nd Charlie that Option #2&#39;s downsides can be mitigated by putting an in=
gestor in front of the helper. Though it&#39;s important to note that this =
probably cancels out the egress savings, since the ingestor (itself a servi=
ce provider) is proxying report shares to the helper. To get the egress sav=
ings, the helper would have to be able to absorb traffic from clients direc=
tly.</div><div><br></div><div>Either way, Option #2 will require extra prot=
ocol logic to deal with synchronization of the report shares. The current s=
pec has each aggregate sub-request from the leader specify the nonce of the=
 report being aggregated. To respond, the helper will first check if it has=
 a report share with that nonce. If not (perhaps the report share hasn&#39;=
t arrived), it&#39;ll have to treat this as an error. Where things get tric=
ky is what the leader does next:<br></div><div>1. It might retry later, but=
 there&#39;s no guarantee that the helper&#39;s report share will ever arri=
ve. Eventually it&#39;ll have to give up and the report will get dropped.<b=
r></div><div>2. It might just ignore the report. If the helper&#39;s report=
 share ever shows up, it might end up storing it indefinitely, waiting for =
the leader to send a request that never arrives.</div><div><br></div><div>T=
his suggests that some data loss may be inevitable in some deployments of O=
ption #2. In my own opinion, this data loss is acceptable.<br></div><div><b=
r></div><div>Chris P.<br></div><div><br></div><div><br></div><div><br></div=
><div><br></div><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" c=
lass=3D"gmail_attr">On Thu, Jan 20, 2022 at 9:25 AM Phillipp Schoppmann &lt=
;<a href=3D"mailto:schoppmann@google.com" rel=3D"noreferrer noreferrer nore=
ferrer noreferrer noreferrer" target=3D"_blank">schoppmann@google.com</a>&g=
t; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0p=
x 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div d=
ir=3D"ltr">One more note on the share sizes: A lot of overhead (~50%) in ou=
r implementation comes from the way I used protocol buffers for serializati=
on.<div>I think there is some optimization potential here, so the cost coul=
d be about half that if we use a format that just sends the raw bytes neede=
d.</div><div>For example, with 128 bit inputs, the theoretical lower bound =
of the cost is just 4128 bytes (again, excluding the verification part).</d=
iv></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_att=
r">On Thu, Jan 20, 2022 at 6:19 PM Phillipp Schoppmann &lt;<a href=3D"mailt=
o:schoppmann@google.com" rel=3D"noreferrer noreferrer noreferrer noreferrer=
 noreferrer" target=3D"_blank">schoppmann@google.com</a>&gt; wrote:<br></di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Hi eve=
ryone,<div><br></div><div>I&#39;m just reading through the discussion and c=
atching up.=C2=A0</div><div><a class=3D"gmail_plusreply" id=3D"gmail-m_7869=
035095328403462gmail-m_-8191592490242430118m_-6258409472252836428m_31176313=
93847772887m_-923516093214407498m_-8313098973848675453m_-280986913471324335=
4gmail-m_-4044173416287307120gmail-m_-5860546362101609079gmail-m_-707565302=
7769385596gmail-m_-7247445031704786523plusReplyChip-0" href=3D"mailto:cpatt=
on@cloudflare.com" rel=3D"noreferrer noreferrer noreferrer noreferrer noref=
errer" target=3D"_blank">@Christopher Patton</a>=C2=A0<a class=3D"gmail_plu=
sreply" id=3D"gmail-m_7869035095328403462gmail-m_-8191592490242430118m_-625=
8409472252836428m_3117631393847772887m_-923516093214407498m_-83130989738486=
75453m_-2809869134713243354gmail-m_-4044173416287307120gmail-m_-58605463621=
01609079gmail-m_-7075653027769385596gmail-m_-7247445031704786523plusReplyCh=
ip-1" href=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer no=
referrer noreferrer noreferrer" target=3D"_blank">@Charlie Harrison</a>=C2=
=A0regarding the sizes of IDPF shares, I did some benchmarks:</div><div><ul=
><li>32 bit inputs -&gt; 2127 byte shares</li><li>64 bit inputs -&gt; 4283 =
byte shares</li><li>128 bit inputs -&gt; 8539 byte shares</li></ul></div><d=
iv>These are for a DPF that allows evaluation at every bit of the input (li=
ke Poplar), but currently without the VDAF part. I am in the process of imp=
lementing verification for our IDPF implementation, and can tell you more o=
nce that&#39;s done.</div><div><br></div><div>Best,</div><div>Phillipp</div=
><div><br></div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Thu, Jan 20, 2022 at 5:04 PM Charlie Harrison &lt;<a hre=
f=3D"mailto:csharrison@google.com" rel=3D"noreferrer noreferrer noreferrer =
noreferrer noreferrer" target=3D"_blank">csharrison@google.com</a>&gt; wrot=
e:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"l=
tr">Thanks Chris,<div>I think I am in agreement with Erik that we could mit=
igate the issues with Option #2 if there is an ingestor client. I think fra=
ming the protocol this way actually allows close imitation=C2=A0of option #=
1 and #2</div><div><br></div><div>- An &quot;option 1&quot; deployment coul=
d=C2=A0just=C2=A0specify one of the aggregators to work as the ingestor eve=
n if it isn&#39;t required by the protocol. They just happen to be the same=
 party.</div><div>- An &quot;option 2&quot; deployment could have the inges=
tor fan out reports to multiple aggregators (or ignore the ingestor complet=
ely)</div><div><br></div><div>Your other points make perfect sense to me!</=
div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_at=
tr">On Wed, Jan 19, 2022 at 7:28 PM Christopher Patton &lt;<a href=3D"mailt=
o:cpatton@cloudflare.com" rel=3D"noreferrer noreferrer noreferrer noreferre=
r noreferrer" target=3D"_blank">cpatton@cloudflare.com</a>&gt; wrote:<br></=
div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><br>=
=C2=A0<div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D=
"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-le=
ft:1ex"><div dir=3D"ltr"><div></div><div>Great! My initial read was that it=
 wouldn&#39;t currently enable it, and I&#39;m happy to work on an issue/PR=
 that can make that more clear and enumerate various deployment=C2=A0scenar=
ios, as you suggest. One point of confusion was the suggestion about limiti=
ng reports from a single IP address, as I&#39;d expect an intermediary serv=
er between the client and leader/aggregators to likely have a fixed IP addr=
ess.</div></div></blockquote><div><br></div><div>Awesome! Let&#39;s start w=
ith opening an issue and folks to start putting their deployment setups the=
re.</div><div><br></div><div>Yes, rate-limiting clients based on IP alone i=
s impossible for aggregators if they&#39;re behind a proxy. It should be no=
ted that this is a heuristic only that an attacker may be able to defeat in=
 many ways. To deal with with Sybil attacks in general, we&#39;ll definitel=
y have to do a bit more anyway.<br></div>=C2=A0</div><div class=3D"gmail_qu=
ote"><br></div><div class=3D"gmail_quote">Chris P.<br></div></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" rel=3D"noreferrer noreferrer noreferrer nor=
eferrer noreferrer" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer nor=
eferrer noreferrer noreferrer noreferrer noreferrer" target=3D"_blank">http=
s://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div></div></div>
</blockquote></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--000000000000fe25ff05d621d1be--


From nobody Fri Jan 21 18:31:44 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 6BE7A3A1D15 for <ppm@ietfa.amsl.com>; Fri, 21 Jan 2022 18:31:43 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id RKGNMpJMdgAT for <ppm@ietfa.amsl.com>; Fri, 21 Jan 2022 18:31:39 -0800 (PST)
Received: from mail-pg1-x530.google.com (mail-pg1-x530.google.com [IPv6:2607:f8b0:4864:20::530]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 358663A1D14 for <ppm@ietf.org>; Fri, 21 Jan 2022 18:31:39 -0800 (PST)
Received: by mail-pg1-x530.google.com with SMTP id 133so9681165pgb.0 for <ppm@ietf.org>; Fri, 21 Jan 2022 18:31:39 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=WS6Wm+7KYQSXatMell/AhnxPFhSBxKRNBLwelRRcsDM=; b=MEGTPRJEPErPpnTxpsqPOrSznCtgAGjDASETwOOt4EoS5Zo+GgKmuN23FjcMkkur8t 2MLAFfRz7J2JB+FS1PyWUwtm7XAN7lsUCn5blgTqH7/3JPllYLZ+uxGi3qHKgfxzVIcj tKn1eOO/48Om210yuSUyslJLOoxvdJmrKOFfA=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=WS6Wm+7KYQSXatMell/AhnxPFhSBxKRNBLwelRRcsDM=; b=zKZ8DEz+GZwMgEOpB26sLvdqG2AMKYBfnnkGZuwOeeYipZBBniFmOHM6SnHfBNj+W3 0t43bb8vEtYam2dvqi9wo8YPqjx/WlVqchuB866p+Puo2VQQJiVaB2UKtbdphP1/KvR5 2NJl0LDifFrds7XhqGqHUtrvGalj/ICDufZeZh7nFucG7DiNAE7h/F4jd+0ikp2aQ33c FrNbGyFohKkK5LjdkQpisqXcvpT4lHV/r4C1/OUK9R6/47c9U6AlmlfNFfBF3IhIFl7n 4LGTew9b70XMf2+gkN6DRGfRX25saYnk+Nw0bV4N7HwRy5NzwVSzzQGsxo9GTgssnfmM rFnA==
X-Gm-Message-State: AOAM533RCUd9GJDyX2BHF9YcPKogqUO3jNhudS0auk06otxsJBeIXFnk UMkhZRQu+VFs7HvjBNx4HCHIZQtXHLDC10p9OqmLsg==
X-Google-Smtp-Source: ABdhPJxd+Y6pdXWupsNSIEL3XSxYjOwGJaRShHNRaJeJquLbk9jqcPZ99XireV/MIEFo55JCHGnSY56aBNQVfxH3dwY=
X-Received: by 2002:a63:b407:: with SMTP id s7mr4823023pgf.202.1642818697679;  Fri, 21 Jan 2022 18:31:37 -0800 (PST)
MIME-Version: 1.0
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com> <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com> <CANzuiCjDBSaikQJJD39-EZpwSaedue7xq1dZiCJ3ciMruEbdhQ@mail.gmail.com> <CABN231rmmXmaSDxhXwX2TG+DVqDjHVWkoVVxUZsC2oQOv8X20g@mail.gmail.com>
In-Reply-To: <CABN231rmmXmaSDxhXwX2TG+DVqDjHVWkoVVxUZsC2oQOv8X20g@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Fri, 21 Jan 2022 18:31:26 -0800
Message-ID: <CAG2Zi232a4RQp-bq=Tk6YptOLne-QPi6Osnsfg0j+AMku6skJQ@mail.gmail.com>
To: Tim Geoghegan <timg@letsencrypt.org>
Cc: Erik Taubeneck <erik.taubeneck@gmail.com>, Watson Ladd <watsonbladd@gmail.com>, ppm@ietf.org,  Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>,  Charlie Harrison <csharrison@google.com>
Content-Type: multipart/alternative; boundary="000000000000e858be05d6228804"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/1XIEStvuCUyhsKsHERvsf2uqtck>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sat, 22 Jan 2022 02:31:43 -0000

--000000000000e858be05d6228804
Content-Type: text/plain; charset="UTF-8"

Thanks, Tim!



> A few of us were batting this around earlier and concluded that the
> problem of aggregators having disjoint views of what shares exist occurs no
> matter what: even in the single upload model, we would expect that
> aggregators will implement some kind of retention policy for submitted
> reports to stop their storage costs from growing indefinitely. So maybe a
> helper discards unaggregated shares after 14 days but its leader does so
> after 21. Even if they use the same retention policy, we can't guarantee
> that they will discard shares at the same moment. If a collector takes long
> enough to issue a `CollectReq` (in some VDAFs like `poplar1`, the aggregate
> phase cannot begin until the collector supplies the aggregation parameter,
> i.e., the candidate prefixes), then we are in the same error case that is
> introduced by partial success of a split upload. That means that the PPM
> protocol has to handle this error case no matter what (and we believe it
> already does).
>

If I may try to summarize: the main downside of Option #2 is new failure
modes, but similar failure modes already exist, and we'll have to deal with
them in much the same way. I'm inclined to agree, although there is still a
possibility of solving the existing problem in a different way. I'll note
that this question is currently tracked by
https://github.com/abetterinternet/ppm-specification/issues/180. That said,
I'm inclined to agree with your analysis here.

My preference would be to take Option #2.

Hope y'all have a good weekend,
Chris P.

--000000000000e858be05d6228804
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><div>Thanks, Tim!</div><div><br></div></d=
iv><div class=3D"gmail_quote"><div></div><div>=C2=A0</div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div></div><div>A few of =
us were batting this around earlier and concluded that the problem of aggre=
gators having disjoint views of what shares exist occurs no matter what: ev=
en in the single upload model, we would expect that aggregators will implem=
ent some kind of retention policy for submitted reports to stop their stora=
ge costs from growing indefinitely. So maybe a helper discards unaggregated=
 shares after 14 days but its leader does so after 21. Even if they use the=
 same retention policy, we can&#39;t guarantee that they will discard share=
s at the same moment. If a collector takes long enough to issue a `CollectR=
eq` (in some VDAFs like `poplar1`, the aggregate phase cannot begin until t=
he collector supplies the aggregation parameter, i.e., the candidate prefix=
es), then we are in the same error case that is introduced by partial succe=
ss of a split upload. That means that the PPM protocol has to handle this e=
rror case no matter what (and we believe it already does).<br></div></div><=
/blockquote><div><br></div><div>If I may try to summarize: the main downsid=
e of Option #2 is new failure modes, but similar failure modes already exis=
t, and we&#39;ll have to deal with them in much the same way. I&#39;m incli=
ned to agree, although there is still a possibility of solving the existing=
 problem in a different way. I&#39;ll note that this question is currently =
tracked by <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/180">https://github.com/abetterinternet/ppm-specification/issues/180=
</a>. That said, I&#39;m inclined to agree with your analysis here.<br></di=
v><div><br></div><div>My preference would be to take Option #2.</div><div><=
br></div><div>Hope y&#39;all have a good weekend,</div><div>Chris P.<br></d=
iv><div><br></div><div><br></div></div></div>

--000000000000e858be05d6228804--


From nobody Sun Jan 23 00:36:13 2022
Return-Path: <do_not_reply@mnot.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 02FB93A0899 for <ppm@ietfa.amsl.com>; Sun, 23 Jan 2022 00:36:12 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.697
X-Spam-Level: 
X-Spam-Status: No, score=-1.697 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_INVALID=0.1, DKIM_SIGNED=0.1, HTML_MESSAGE=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=no autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=fail (2048-bit key) reason="fail (message has been altered)" header.d=mnot.net header.b=jfbUB+IX; dkim=fail (2048-bit key) reason="fail (message has been altered)" header.d=messagingengine.com header.b=koRY9NjX
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id g5Osyrjr0Rd2 for <ppm@ietfa.amsl.com>; Sun, 23 Jan 2022 00:36:07 -0800 (PST)
Received: from wout5-smtp.messagingengine.com (wout5-smtp.messagingengine.com [64.147.123.21]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 4EE9D3A089A for <ppm@ietf.org>; Sun, 23 Jan 2022 00:36:07 -0800 (PST)
Received: from compute1.internal (compute1.nyi.internal [10.202.2.41]) by mailout.west.internal (Postfix) with ESMTP id DDA0D3200C6B for <ppm@ietf.org>; Sun, 23 Jan 2022 02:39:25 -0500 (EST)
Received: from mailfrontend1 ([10.202.2.162]) by compute1.internal (MEProxy); Sun, 23 Jan 2022 02:39:26 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mnot.net; h=cc :content-type:date:from:from:in-reply-to:mime-version:reply-to :sender:subject:subject:to:to; s=fm1; bh=zs8R8p9LFX31AnPbpDElLyg SE0++S0ow4ezz4jpGuHM=; b=jfbUB+IXFPNTZd1cQXAqBH0/GY8RmZd7E3Q5A1x /zWW6pbsr3Kt9gEUmMolBpiAlfGZkiGztgyfFDWwFdmQ4zNi+S4oOCZX6JJUWCyn /YwLkz1P8xYy5BR8VVrkjnKlAHApL0HHmK7rzwskmEbkPIBXdgnvJMGd+yx6xmMm uym6kaSKZICaE0k9MqYk4pOm1eoHsO+8iqRhu7eeZunUuewiTBxyp6Of5xYKB1tW o6Hs+xJRQ9giTJTibRBDlUg0XrI6C2wFbaz7unt1EfXCFvEBS09nX8hZPNPwLzOM 8j2t3BiJRzb5z97GvOADO2dvf1+oiftP9Hy8Zm/oJESKR4Q==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-type:date:from:from :in-reply-to:mime-version:reply-to:sender:subject:subject:to:to :x-me-proxy:x-me-proxy:x-me-sender:x-me-sender:x-sasl-enc; s= fm1; bh=zs8R8p9LFX31AnPbpDElLygSE0++S0ow4ezz4jpGuHM=; b=koRY9NjX 0yeuflBlVasKDWzGWxXQViQDbiK2fRigic6KbZdPN6xWByE5GURqOhU0rK4vjyxH YkKEJpIznmtQVZcO9oxN71Su35MFdUVaOgGrnKJKrgnL3apeFfLBcvjTLjl7RWuP +OT9g0HRzevFXRx/UfzVdaM0yZ7BZqo4A3xtXmfLk9/AGAENPqF9iut+z3H9xgtv YYXoyrplzrRAlIoSEak9F+S1O4tOdrfVlmeEtfLiJbaH6IcGNVOJQPSKljVatw4w 7NfY84sjpposwtI6euaE9udhtYyAkil4l0xizrhEga+dIbwSlNOXdYC+v7uWkuQr suvuyJ1YxdWd+Q==
X-ME-Sender: <xms:LQbtYQYe0auYrp4PGuZeTWp2brVKpg1jg8wxWpknSLCKYxPboK02Rg> <xme:LQbtYbZIgTJ0eQTVFugsI_aiXCP3-MxkjldB9rvg0VmFeW7Z0psmzFkzNCbl6TcPd oEqIUf17DK2Aeu05w>
X-ME-Received: <xmr:LQbtYa_Rh3v_wx-DI9p2UI8W4AkTuoW-KBmHKCpOMFnoqNrO9_cWT3Scto9hhSlonbJ0AWmgNDtRv1GNQ77N3gnJdA97Cph19kSVT88SAA71Vw780ONSPdBEwjyoec3nYjSZzg>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrvdefgdduudduucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucfpohcuuggrthgvuchfihgvlhguucdlgeelmdenuc fjughrpegtggfhvffusegrtddtredttdejnecuhfhrohhmpeftvghpohhsihhtohhrhicu tegtthhivhhithihucfuuhhmmhgrrhihuceuohhtuceoughopghnohhtpghrvghplhihse hmnhhothdrnhgvtheqnecuggftrfgrthhtvghrnhepkeefvdduteejvdefkeehieevuefg fefhteetveegffekffefteffvdelheduieetnecuffhomhgrihhnpehgihhthhhusgdrtg homhenucevlhhushhtvghrufhiiigvpedunecurfgrrhgrmhepmhgrihhlfhhrohhmpegu ohgpnhhothgprhgvphhlhiesmhhnohhtrdhnvght
X-ME-Proxy: <xmx:LQbtYartbvPc8O67Qc9OHpPfiFP9kOB7tAMuSD80mU7a3tdZ7ayQ-Q> <xmx:LQbtYbqxvRDluc3tG9UrkK-9zESPGHIMkDxIYXo3zNb6B4YWnfD21w> <xmx:LQbtYYRk0DKmiIuCkHHP4ZWj4ZvO-NJ0tp7RG2Na60rC4GBJmVfuOw> <xmx:LQbtYS18g-xTnqzfv_HARMWj4XV5xuLZJgyTrolfsFFOLv0FMTN-OQ>
Received: by mail.messagingengine.com (Postfix) with ESMTPA for <ppm@ietf.org>;  Sun, 23 Jan 2022 02:39:25 -0500 (EST)
Content-Type: multipart/alternative; boundary="===============4329395675999378334=="
MIME-Version: 1.0
From: Repository Activity Summary Bot <do_not_reply@mnot.net>
To: ppm@ietf.org
Message-Id: <20220123083607.4EE9D3A089A@ietfa.amsl.com>
Date: Sun, 23 Jan 2022 00:36:07 -0800 (PST)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/YFj399CwSwuwEc42nDO4nXlvVqc>
Subject: [Ppm] Weekly github digest (Privacy Preserving Measurement Activity Summary)
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 23 Jan 2022 08:36:12 -0000

--===============4329395675999378334==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"; format="flowed"




Events=20

Issues
------
* abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC18)
  2 issues created:
  - Replace helper state blob with an aggregation job UUID (by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/185=20
  - Interaction of anti-replay and flexibility (by ekr)
    https://github.com/abetterinternet/ppm-specification/issues/183=20

  5 issues received 18 new comments:
  - #185 Replace helper state blob with an aggregation job UUID (1 by tgeog=
hegan)
    https://github.com/abetterinternet/ppm-specification/issues/185=20
  - #183 Interaction of anti-replay and flexibility (5 by chris-wood, cjpat=
ton, ekr)
    https://github.com/abetterinternet/ppm-specification/issues/183=20
  - #166 Consider making the Leader a non-aggregator (9 by cjpatton, csharr=
ison, eriktaubeneck, tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/166=20
  - #146 Aggregators are forced to use a single HPKE config for all tasks (=
1 by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/146 [parkin=
g-lot]=20
  - #111 Do something about long-running requests (2 by cjpatton, tgeoghega=
n)
    https://github.com/abetterinternet/ppm-specification/issues/111=20

  1 issues closed:
  - Consider making the Leader a non-aggregator https://github.com/abetteri=
nternet/ppm-specification/issues/166=20



Pull requests
-------------
* abetterinternet/ppm-specification (+4/-3/=F0=9F=92=AC3)
  4 pull requests submitted:
  - Remove .targets.mk (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/186=20
  - Redefine `Interval` to invalid values impossible (by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/pull/184=20
  - Convert [..]-style refs to {{..}} (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/182=20
  - Align some terminology with the VDAF draft (by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/181=20

  2 pull requests received 3 new comments:
  - #182 Convert [..]-style refs to {{..}} (1 by ekr)
    https://github.com/abetterinternet/ppm-specification/pull/182=20
  - #181 Align some terminology with the VDAF draft (2 by cjpatton)
    https://github.com/abetterinternet/ppm-specification/pull/181=20

  3 pull requests merged:
  - Redefine `Interval` to invalid values impossible
    https://github.com/abetterinternet/ppm-specification/pull/184=20
  - Align some terminology with the VDAF draft
    https://github.com/abetterinternet/ppm-specification/pull/181=20
  - Convert [..]-style refs to {{..}}
    https://github.com/abetterinternet/ppm-specification/pull/182=20


Repositories tracked by this digest:
-----------------------------------
* https://github.com/abetterinternet/ppm-specification

--===============4329395675999378334==
Content-Type: text/html; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html lang=3D"en">
<head>
<meta charset=3D"utf-8">
<title>Weekly github digest (Privacy Preserving Measurement Activity Summar=
y)</title>
<style>
body { font-family: Gotham, "Helvetica Neue", Helvetica, Arial, sans-serif;=
 font-size: 14px; }
h2 { margin-top: 3em; color: #A52A2A; font-style: italic; font-weight: norm=
al; }
h3 { margin-bottom:0; margin-top: 2em; font-size: 1.2em; }
h1+h2 { margin-top: 1em; }
a { color: #bb6219; text-decoration: none; }
li { margin-bottom: .35em; }
.repos { margin-bottom: 0; margin-top:0; line-height: 1.2; }
.new { color: red; }
.label { display: inline;
	padding: .2em .6em .3em;
	font-size: 75%;
	font-weight: 700;
	line-height: 1;
	color: #fff;
	text-align: center;
	white-space: nowrap;
	vertical-align: baseline;
	border-radius: .25em;
}
details {
	margin-top: 8em;
	}
summary {
	margin-bottom: 1em;
	cursor: pointer;
}
</style>
</head>

<body>
<h1>Sunday January 23, 2022</h1>

<p>Events </p>

<h2>Issues</h2>

<h3>abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC18)</h3>
  <p class=3D"new">2 issues created:</p>
  <ul>
  <li>#185 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/185">Replace helper state blob with an aggregation job UUID</a> (by =
tgeoghegan) </li>
 =20
  <li>#183 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/183">Interaction of anti-replay and flexibility</a> (by ekr) </li>
  </ul>

  <p>5 issues received 18 new comments:</p>
  <ul>
  <li>#185 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/185">Replace helper state blob with an aggregation job UUID</a> (1 b=
y tgeoghegan) </li>
 =20
  <li>#183 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/183">Interaction of anti-replay and flexibility</a> (5 by chris-wood=
, cjpatton, ekr) </li>
 =20
  <li>#166 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/166">Consider making the Leader a non-aggregator</a> (9 by cjpatton,=
 csharrison, eriktaubeneck, tgeoghegan) </li>
 =20
  <li>#146 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/146">Aggregators are forced to use a single HPKE config for all task=
s</a> (1 by tgeoghegan) <span class=3D"label" style=3D"background-color: #c=
fd3d7; color: #000000">parking-lot</span> </li>
 =20
  <li>#111 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/111">Do something about long-running requests</a> (2 by cjpatton, tg=
eoghegan) </li>
  </ul>

  <p>1 issues closed:</p>
  <ul>
  <li>#166 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/166">Consider making the Leader a non-aggregator</a> </li>
  </ul>



<h2>Pull requests</h2>
<h3>abetterinternet/ppm-specification (+4/-3/=F0=9F=92=AC3)</h3>
  <p class=3D"new">4 pull requests submitted:</p>
  <ul>
  <li>#186 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/186">Remove .targets.mk</a> (by cjpatton) </li>
 =20
  <li>#184 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/184">Redefine `Interval` to invalid values impossible</a> (by tgeogheg=
an) </li>
 =20
  <li>#182 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/182">Convert [..]-style refs to {{..}}</a> (by cjpatton) </li>
 =20
  <li>#181 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/181">Align some terminology with the VDAF draft</a> (by cjpatton) </li>
  </ul>

  <p>2 pull requests received 3 new comments:</p>
  <ul>
  <li>#182 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/182">Convert [..]-style refs to {{..}}</a> (1 by ekr) </li>
 =20
  <li>#181 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/181">Align some terminology with the VDAF draft</a> (2 by cjpatton) </=
li>
  </ul>

  <p>3 pull requests merged:</p>
  <ul>
  <li>#184 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/184">Redefine `Interval` to invalid values impossible</a> </li>
 =20
  <li>#181 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/181">Align some terminology with the VDAF draft</a> </li>
 =20
  <li>#182 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/182">Convert [..]-style refs to {{..}}</a> </li>
  </ul>


  <details>
    <summary>Repositories tracked by this digest:</summary>
<ul class=3D"repos">
  <li><a href=3D"https://github.com/abetterinternet/ppm-specification">http=
s://github.com/abetterinternet/ppm-specification</a></li>
</ul>
</details>
</body>
</html>

--===============4329395675999378334==--


From nobody Mon Jan 24 18:30:24 2022
Return-Path: <timg@letsencrypt.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 3B4AD3A0C75 for <ppm@ietfa.amsl.com>; Mon, 24 Jan 2022 18:30:22 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.673
X-Spam-Level: 
X-Spam-Status: No, score=-2.673 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.576, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=letsencrypt.org
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ha8rAGgtDybe for <ppm@ietfa.amsl.com>; Mon, 24 Jan 2022 18:30:17 -0800 (PST)
Received: from mail-qv1-xf34.google.com (mail-qv1-xf34.google.com [IPv6:2607:f8b0:4864:20::f34]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id A834E3A0C73 for <ppm@ietf.org>; Mon, 24 Jan 2022 18:30:17 -0800 (PST)
Received: by mail-qv1-xf34.google.com with SMTP id b12so3662526qvz.5 for <ppm@ietf.org>; Mon, 24 Jan 2022 18:30:17 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=letsencrypt.org; s=google; h=mime-version:from:date:message-id:subject:to; bh=T/lgxCD2ORt+pe3XMZUibAZtG33LrcUi3HQ/1c7ks0k=; b=bdVcgptJVGkHZ5KrtXLGg4TQNkRin5zSDzGMJ5IpJm3Kdpxj0q5aTORKVxKYcnZ7Kz WKUgeck9AHx4kfsAJnooL8aKA3I7SayBZIKYa0r82kNqxun4+EBv7PGLXBM8QCIGkiSp 19RaUbHX3j5+Z7I9Gl1kdvbACW+R7k8ZJ4h9U=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:from:date:message-id:subject:to; bh=T/lgxCD2ORt+pe3XMZUibAZtG33LrcUi3HQ/1c7ks0k=; b=mIH9yrcWCrHJJ3xbJ+JW2WjGBEOplBd/T4WC2QGOOyA6okIu2M8KLgamDWH3HEG0rZ 96l10YjadqT6cwAq8IBQKdBWjw6YvWs1lhKWEJG1fUbe0OZSHF3h5u+ppGD7nk53GEZp cV/XQlEmsf7ACUMf48aB07NCUIgyKKQ7fjGbPb4Aqn74UfcMcAyAEhvKyrMW+XtcgrHh sQrC7Q8L5AARxP2Xpghf8zA1+UIsPNAsLYz5Cge2L0KxcIdL4wrkTNH53zA1ov8zM9T1 JwE6sLEps38B4v4Dk/fYvpmEiQo5tojTO6Umorg9L5xJZ2VzPLy/WbrWl2UabA4R7Vy3 FseA==
X-Gm-Message-State: AOAM532wbe92FL6/Ec+5OFUU1FyQ1fgklzfoSq2xYpNxkDL+sYx6o3d3 bkU7vAOKKbHTms+viiJWDkP6UUKhJ6/2SBV25DLOM65LMfTXqQ==
X-Google-Smtp-Source: ABdhPJzIqK2BCm6LYnFdkcltAX2uD1/ReRvDkiWeU/T7veUzruXo29HWsMC2ZIt/aXDeJDNp8lFrJXvXpINrZJPT3HU=
X-Received: by 2002:a05:6214:aca:: with SMTP id g10mr11544602qvi.61.1643077815505;  Mon, 24 Jan 2022 18:30:15 -0800 (PST)
MIME-Version: 1.0
From: Tim Geoghegan <timg@letsencrypt.org>
Date: Mon, 24 Jan 2022 18:30:04 -0800
Message-ID: <CABN231pz5Out1aCpxEcJvxnDWWQ6WsKUzrtMu1VCRe10K4owUg@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000088268905d65edd1e"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/0PwVvmlvYocpJKypfaorT8eA9q0>
Subject: [Ppm] Making long-running requests asynchronous
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 25 Jan 2022 02:30:22 -0000

--00000000000088268905d65edd1e
Content-Type: text/plain; charset="UTF-8"

Hello ppm,

The design of PPM has thus far assumed that the interactions between
participants will be synchronous HTTP transactions. We've suspected for a
while that this probably won't work in realistic deployments for all
messages[1]. In particular, servicing a Collect request[2] will require the
leader and helper to verify and aggregate every report falling into the
relevant batch interval, which could take quite a lot of distributed
computation by the aggregators, and for VDAFs like `poplar1`, none of that
work can begin until the collector issues the CollectReq. So we wouldn't
want the collector and leader to keep connections open for minutes (maybe
even hours) waiting for an HTTP response from the leader.

To deal with this, I think we will make at least `CollectReq/CollectResp`
into an asynchronous interaction. Roughly, I envision that the Collector
would send a CollectReq to the leader's `collect` endpoint as it does
today, but instead of blocking until it gets an HTTP 200 with a CollectResp
in the body, it would immediately get an HTTP 202 Accepted response whose
body contains a unique identifier (a UUID should work) for the collect job.
The collector would subsequently poll `[leader]/collect/<job identifier>`
at some interval, which would not return HTTP 200 OK until the aggregation
is complete and the CollectResp is ready.

Another option would be for the Collector to listen on some HTTP endpoint
at which it could be notified of completion of collect jobs. The downside
is that this would require collectors to operate a publicly available HTTP
server that listens for requests (I don't think they need to do this yet)
and we would need a way for the leader to authenticate to the collector (to
prevent network attackers from sending collectors bogus aggregates).

I'd love to hear thoughts on what the established best practices for
long-running tasks in HTTP APIs are.

Another problem to chew on is the aggregate portion of the protocol. I feel
certain that collect needs to be asynchronous, but it's not obvious that
the same is true for the aggregation sub-protocol executed between leader
and helper[3]. The reason is that each message in that interaction
references far far fewer reports than any collect message, and the leader
should always be able to carve up execution of the aggregate protocol into
individual jobs that reference few enough reports so that synchronous
handling of aggregate messages should be reasonable. What kind of
heuristics usually inform whether or not to make an API asynchronous?
Should we make these messages async just in case it ends up being useful in
the future?

Thanks,
Tim Geoghegan
ISRG

[1] https://github.com/abetterinternet/ppm-specification/issues/111
[2]
https://abetterinternet.github.io/ppm-specification/draft-gpew-priv-ppm.html#name-collecting-results
[3]
https://abetterinternet.github.io/ppm-specification/draft-gpew-priv-ppm.html#name-verifying-and-aggregating-r

--00000000000088268905d65edd1e
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Hello ppm,</div><div><br></div><div>The design of PPM=
 has thus far assumed that the interactions between participants will be sy=
nchronous HTTP transactions. We&#39;ve suspected for a while that this prob=
ably won&#39;t work in realistic deployments for all messages[1]. In partic=
ular, servicing a Collect request[2] will require the leader and helper to =
verify and aggregate every report falling into the relevant batch interval,=
 which could take quite a lot of distributed computation by the aggregators=
, and for VDAFs like `poplar1`, none of that work can begin until the colle=
ctor issues the CollectReq. So we wouldn&#39;t want the collector and leade=
r to keep connections open for minutes (maybe even hours) waiting for an HT=
TP response from the leader.</div><div><br></div><div>To deal with this, I =
think we will make at least `CollectReq/CollectResp` into an asynchronous i=
nteraction. Roughly, I envision that the Collector would send a CollectReq =
to the leader&#39;s `collect` endpoint as it does today, but instead of blo=
cking until it gets an HTTP 200 with a CollectResp in the body, it would im=
mediately get an HTTP 202 Accepted response whose body contains a unique id=
entifier (a UUID should work) for the collect job. The collector would subs=
equently poll `[leader]/collect/&lt;job identifier&gt;` at some interval, w=
hich would not return HTTP 200 OK until the aggregation is complete and the=
 CollectResp is ready.</div><div><br></div><div>Another option would be for=
 the Collector to listen on some HTTP endpoint at which it could be notifie=
d of completion of collect jobs. The downside is that this would require co=
llectors to operate a publicly available HTTP server that listens for reque=
sts (I don&#39;t think they need to do this yet) and we would need a way fo=
r the leader to authenticate to the collector (to prevent network attackers=
 from sending collectors bogus aggregates).</div><div><br></div><div>I&#39;=
d love to hear thoughts on what the established best practices for long-run=
ning tasks in HTTP APIs are.</div><div><br></div><div>Another problem to ch=
ew on is the aggregate portion of the protocol. I feel certain that collect=
 needs to be asynchronous, but it&#39;s not obvious that the same is true f=
or the aggregation sub-protocol executed between leader and helper[3]. The =
reason is that each message in that interaction references far far fewer re=
ports than any collect message, and the leader should always be able to car=
ve up execution of the aggregate protocol into individual jobs that referen=
ce few enough reports so that synchronous handling of aggregate messages sh=
ould be reasonable. What kind of heuristics usually inform whether or not t=
o make an API asynchronous? Should we make these messages async just in cas=
e it ends up being useful in the future?<br></div><div><br></div><div>Thank=
s,</div><div>Tim Geoghegan</div><div>ISRG</div><div><br></div><div><div><di=
v>[1] <a href=3D"https://github.com/abetterinternet/ppm-specification/issue=
s/111">https://github.com/abetterinternet/ppm-specification/issues/111</a><=
/div></div></div><div>[2] <a href=3D"https://abetterinternet.github.io/ppm-=
specification/draft-gpew-priv-ppm.html#name-collecting-results">https://abe=
tterinternet.github.io/ppm-specification/draft-gpew-priv-ppm.html#name-coll=
ecting-results</a></div><div>[3] <a href=3D"https://abetterinternet.github.=
io/ppm-specification/draft-gpew-priv-ppm.html#name-verifying-and-aggregatin=
g-r">https://abetterinternet.github.io/ppm-specification/draft-gpew-priv-pp=
m.html#name-verifying-and-aggregating-r</a></div></div>

--00000000000088268905d65edd1e--


From nobody Mon Jan 24 19:30:21 2022
Return-Path: <mt@lowentropy.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 4027D3A116B for <ppm@ietfa.amsl.com>; Mon, 24 Jan 2022 19:30:19 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.1
X-Spam-Level: 
X-Spam-Status: No, score=-2.1 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=lowentropy.net header.b=bkTG6FoH; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=FfqasDE5
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id U2dKKgSSYoVY for <ppm@ietfa.amsl.com>; Mon, 24 Jan 2022 19:30:14 -0800 (PST)
Received: from wout4-smtp.messagingengine.com (wout4-smtp.messagingengine.com [64.147.123.20]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 973783A1166 for <ppm@ietf.org>; Mon, 24 Jan 2022 19:30:14 -0800 (PST)
Received: from compute2.internal (compute2.nyi.internal [10.202.2.46]) by mailout.west.internal (Postfix) with ESMTP id AF5453200C4A for <ppm@ietf.org>; Mon, 24 Jan 2022 22:30:11 -0500 (EST)
Received: from imap41 ([10.202.2.91]) by compute2.internal (MEProxy); Mon, 24 Jan 2022 22:30:11 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=lowentropy.net; h=cc:content-type:date:date:from:from:in-reply-to:in-reply-to :message-id:mime-version:references:reply-to:sender:subject :subject:to:to; s=fm1; bh=Di8O9kBkGw+bwZR784lzKku8Kc/u2jZQpkgoJb OiE9U=; b=bkTG6FoHvzCBwhC5ZZQzF26TNBvDOToi6BZPn/680ek5Rd226Cs/3Z 20Im+Az3LMVSuDEd1RdBaC3dS3b7jDdytEey+h0ZOUvDm+0CXN7gHGQ9CmpJDMKY hElv0IrV0wzEPDqbkvQG3HVP12ny6LMnx5Mih+3i01k1UZFnmSVg/8Vdca9u+/1N +5sAKILV9K5twab8JCWR/YTsHJy40UsDN+TkMUrdoRRHLG9NjjRTayimFEIIOeP0 T+j0j4X/gChCPW1R2iogIxxNQQBvd7j1mxonuXBUgw1HmxD4+v7872q+PXL99tmm fcvVCKYjGqRDc9ZMxUi+ckQKEj/JCBHg==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-type:date:date:from:from :in-reply-to:in-reply-to:message-id:mime-version:references :reply-to:sender:subject:subject:to:to:x-me-proxy:x-me-proxy :x-me-sender:x-me-sender:x-sasl-enc; s=fm1; bh=Di8O9kBkGw+bwZR78 4lzKku8Kc/u2jZQpkgoJbOiE9U=; b=FfqasDE56LlfIAqrwLgurOO6+/xQ7Mvhg usMzwQQ4VUXTLGPz+P/g+TsliLQjFxG0+2oBTZDbipTMwV62ZrnagugNzUijAGgG uIIcLO45VdgP/0ZgX21tI2uKcSaUxVCtYOdh+GbIT3YEUS1ktZRR5xy1aLVMO7yU i3WUVleqw5wuWhl+DZLgoG54bQllTft9KXk3R8ch3pb1KeHP/Ff+5JGVu3o6ATbN NYjChsK9XFjpgf0/QK7zZfdcXrEg3RJxz6FsDfbOWsRj606EmIuShlm1+mxmwLHY Abc9m6+UXFtwSZF38YPjILCdxCtIzeoF6xfnDZ7fYe1w8UgMShABw==
X-ME-Sender: <xms:wm7vYbF0-dWJUSAQbhgYdBqdlV4EzsY905-8AV_mFEzve5cuoRjyTA> <xme:wm7vYYXRL5NGpV_MBZmmuo24UM6szDAQ09JKKosW74hIqeFr0JBvSadHJJtsGUhio zqcNFCA9OjfGPqRAbw>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrvdekgdehkecutefuodetggdotefrodftvf curfhrohhfihhlvgemucfhrghsthforghilhdpqfgfvfdpuffrtefokffrpgfnqfghnecu uegrihhlohhuthemuceftddtnecunecujfgurhepofgfggfkjghffffhvffutgesthdtre dtreertdenucfhrhhomhepfdforghrthhinhcuvfhhohhmshhonhdfuceomhhtsehlohif vghnthhrohhphidrnhgvtheqnecuggftrfgrthhtvghrnhepgeehgfelledugfeugfeghf fhheelhfetvdehfeetudduvdfhuefhudeuhfetfeevnecuffhomhgrihhnpehhthhtphif ghdrohhrghdprhhftgdqvgguihhtohhrrdhorhhgnecuvehluhhsthgvrhfuihiivgeptd enucfrrghrrghmpehmrghilhhfrhhomhepmhhtsehlohifvghnthhrohhphidrnhgvth
X-ME-Proxy: <xmx:wm7vYdILlhVR2gjtg0kehzN_hbOCBFz28g-msNZOkmN5JCt8h9sXNg> <xmx:wm7vYZEg_UVzHc-hakag2njlSJ21CcRw5mp4pLDmrcHyRRwFkduC3w> <xmx:wm7vYRXo-DdczXLMMaKCzHwgZRDIgp3IQWdG3HXXhbo8wZ9-lo0HWw> <xmx:w27vYZhkh92OlUFZjCMbQURxU8r1rbrqTuMMITd1xnVcyGgxsZPrMA>
Received: by mailuser.nyi.internal (Postfix, from userid 501) id D51A43C00CF; Mon, 24 Jan 2022 22:30:10 -0500 (EST)
X-Mailer: MessagingEngine.com Webmail Interface
User-Agent: Cyrus-JMAP/3.5.0-alpha0-4585-ga9d9773056-fm-20220113.001-ga9d97730
Mime-Version: 1.0
Message-Id: <07c99f2f-bc2f-4566-90e0-265efa3134d4@beta.fastmail.com>
In-Reply-To: <CABN231pz5Out1aCpxEcJvxnDWWQ6WsKUzrtMu1VCRe10K4owUg@mail.gmail.com>
References: <CABN231pz5Out1aCpxEcJvxnDWWQ6WsKUzrtMu1VCRe10K4owUg@mail.gmail.com>
Date: Tue, 25 Jan 2022 14:29:53 +1100
From: "Martin Thomson" <mt@lowentropy.net>
To: ppm@ietf.org
Content-Type: text/plain
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/dNf3eA-r6qGgeBZ9PhnHjJ1pSEc>
Subject: Re: [Ppm] Making long-running requests asynchronous
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 25 Jan 2022 03:30:20 -0000

Hi Tim,


On Tue, Jan 25, 2022, at 13:30, Tim Geoghegan wrote:
> I'd love to hear thoughts on what the established best practices for 
> long-running tasks in HTTP APIs are.

Maybe this might help you: https://httpwg.org/specs/rfc7240.html#respond-async  (URIs, not UUIDs)

You might also find this informative: https://www.rfc-editor.org/rfc/rfc8820.html

You are probably familiar with how ACME does asynchronous processing, which is largely OK.

> What kind of heuristics usually inform whether or not to 
> make an API asynchronous? Should we make these messages async just in 
> case it ends up being useful in the future?

This probably comes down to preference at some level.  However, asynchronous interactions are a fair bit more complex than synchronous ones, so avoiding that complexity is good.  Keep in mind that modern HTTP versions aren't stalled if there are a handful of long-running requests, so unless you are talking a fairly significant time, it might be cleaner and easier to just have the client wait on hold.

Asynchronous requests do come with some benefits to go with the complexity. For instance, robustness can be improved if the initial request creates resources that can be referred to later on a different connection without having to rely on the connection remaining viable.

RFC 7240 does suggest that you might offer both, but I would only go that far if you thought that implementations would consistently use both options.


From nobody Tue Jan 25 11:49:34 2022
Return-Path: <caw@heapingbits.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 7FF083A091D for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 11:49:25 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.098
X-Spam-Level: 
X-Spam-Status: No, score=-2.098 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=heapingbits.net header.b=pQUnxisG; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=Dc5HwCIQ
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id tQ_izRkYh5Qp for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 11:49:20 -0800 (PST)
Received: from out2-smtp.messagingengine.com (out2-smtp.messagingengine.com [66.111.4.26]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 808BB3A09EF for <ppm@ietf.org>; Tue, 25 Jan 2022 11:49:13 -0800 (PST)
Received: from compute4.internal (compute4.nyi.internal [10.202.2.44]) by mailout.nyi.internal (Postfix) with ESMTP id 7DB515C0228; Tue, 25 Jan 2022 14:49:12 -0500 (EST)
Received: from mailfrontend1 ([10.202.2.162]) by compute4.internal (MEProxy); Tue, 25 Jan 2022 14:49:12 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=heapingbits.net; h=cc:cc:content-transfer-encoding:content-type:date:date:from :from:in-reply-to:in-reply-to:message-id:mime-version:references :reply-to:sender:subject:subject:to:to; s=fm3; bh=Pk4dcuxXe1MLRv 3bq2F4wATgaVQfhPUczERonlDD/Yk=; b=pQUnxisGUpsODD4MivSE83HYpQkqeA CNwczQkyQ4Tc7k0UusDBeSf3HfFcIE/TR24yIUJbwH1+3tBKCANN0rieset5rO9p 7QU3rBDxgGPq2jsr2Eq6EIrljGhdjaNWoQVeLXeyZqafjcNMDr93Nfds/1AyMSPp Plo9VqfgxYU4giarRt60OkdgoMRMtL1a92eu+SD4Nj43ZuBSb6o6Ff08NHW9ZCn+ VdqqnNpVsNhhpEZKIlas9gLPw40A2gxZlQTwBXQclO85q6dtXpitoscZHfBZUA08 S7k0DXMEXkp3U5jRwTJRPApQQvlbOmZoKXSmb0eNiyvZ9Kgl7dxHMXeA==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:cc:content-transfer-encoding :content-type:date:date:from:from:in-reply-to:in-reply-to :message-id:mime-version:references:reply-to:sender:subject :subject:to:to:x-me-proxy:x-me-proxy:x-me-sender:x-me-sender :x-sasl-enc; s=fm1; bh=Pk4dcuxXe1MLRv3bq2F4wATgaVQfhPUczERonlDD/ Yk=; b=Dc5HwCIQiqk59FhPavaXiWoC1h0B2CjfgknBlh3D02x8aTXJd8eDXHZeO C4y1+Sofpl4rzu2iOFG/82+D9/uwMseA6OxquTaxvpObl2Kji+VVgSeYzs1zCu3f HkjJeCklpqsg/DnMVY7NTWCeTpgz1PO42AyXmgOcxgxCcOJfLDEIa9O1hnRpW4rP TNPK67hEB3OCga8g2TyN3SwgoRx65R4FWVOyukl11jr7OJhTwgRn6jRaFBe3Bh59 PUKv3rlmmIO8x/vmBPK3bchE6g0r8RfPA9YRDhXMog7WJp1NEOHzuzANIIDf0YaR I8zMfNNyZoKAxETIzWW8q2J9a6Nlw==
X-ME-Sender: <xms:OFTwYXgfDcmbzLz0JsakVJHKlLh7yS5MJ4RLcJ-ULzYUR3C_8bHC8Q> <xme:OFTwYUBScr-Khmx-pDNDgvuHa_WuT1jERE6bXaE6WgGDptbTp76yrn-IlmqRNa-FX bi8-6oCYBQGp--n7_A>
X-ME-Received: <xmr:OFTwYXHd8oPR8kzBHwNHBdZVXOW7YDbjUh9iGrMbIeanSSWhqXUd0ZYDqrKW-FdK3p3twZY2POHq1tVnBB3F0VPsbBnuNq1wPQ>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrvdelgdduvdekucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucesvcftvggtihhpihgvnhhtshculddquddttddmne cujfgurheptggguffhjgffgffkfhfvofesthhqmhdthhdtvdenucfhrhhomhepvehhrhhi shhtohhphhgvrhcuhghoohguuceotggrfieshhgvrghpihhnghgsihhtshdrnhgvtheqne cuggftrfgrthhtvghrnhepjefgheetffejteeifeeiheelhefggeeggeeggfegueejteeh uedtgfdvleegffffnecuffhomhgrihhnpehgihhthhhusgdrtghomhdpihgvthhfrdhorh hgnecuvehluhhsthgvrhfuihiivgeptdenucfrrghrrghmpehmrghilhhfrhhomheptggr fieshhgvrghpihhnghgsihhtshdrnhgvth
X-ME-Proxy: <xmx:OFTwYUQ7a8i8S6T04xKjkePj_nB-PqQlPPxQAwir-9617LyK0_tAjg> <xmx:OFTwYUwY24PO1IaurnpzhZgP_Iz85VUuA2vGGl51Nqh6bu3ctp8XMg> <xmx:OFTwYa48FMucJN6sdJF83Og2H2NDHYoSp5mX-ZkTxVbAtHJpMUOnkQ> <xmx:OFTwYblSYRTPUMjI7h2NwiX6GRVcig9B7aPPPoQ1EQ3KwUHoLThL4A>
Received: by mail.messagingengine.com (Postfix) with ESMTPA; Tue, 25 Jan 2022 14:49:10 -0500 (EST)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 15.0 \(3693.40.0.1.81\))
From: Christopher Wood <caw@heapingbits.net>
In-Reply-To: <CAG2Zi232a4RQp-bq=Tk6YptOLne-QPi6Osnsfg0j+AMku6skJQ@mail.gmail.com>
Date: Tue, 25 Jan 2022 11:49:08 -0800
Cc: Tim Geoghegan <timg@letsencrypt.org>, Erik Taubeneck <erik.taubeneck@gmail.com>, Watson Ladd <watsonbladd@gmail.com>, Phillipp Schoppmann <schoppmann@google.com>, Robert Kubis <rkubis@google.com>, ppm@ietf.org, Charlie Harrison <csharrison@google.com>
Content-Transfer-Encoding: quoted-printable
Message-Id: <74B6D855-8E52-4E95-9C0B-2ED827C61172@heapingbits.net>
References: <CAG2Zi219yg7bfgDrRUUq++aVwm8GKu5hLxJOZU0oSuu50Ubbkw@mail.gmail.com> <CADjAqN6E3oFt97UQrBKH8ty47Z5k9Dnj=3sxhoio54o7rEC1xQ@mail.gmail.com> <CANzuiCjYQcSkaTiZYAkkcJ4QpGPxOBpdza2LjCpfn+bDrKzCAA@mail.gmail.com> <CAG2Zi20jO31zsBdjOLaVLcVzdr-mAaeeoyHQC7rH8V_WSRvYiw@mail.gmail.com> <CANzuiCjVh9LnDO+6XvWmYbpBOX_MFbtuPNY34VHiQUSxCW2zhA@mail.gmail.com> <CAG2Zi2274cRrukrMR6vE4UAT1TwHjucJRqEs+vKb754Ezht4_A@mail.gmail.com> <CANzuiCjBhKoD5d+Z1J0QxH6+Aw01D6btfFd0HR4jNxi-ETBkXg@mail.gmail.com> <CAG2Zi23UDrQ=k-1J=_tfmghLdXSR93K8GSaQX=PS6xsWbOxNGA@mail.gmail.com> <CADjAqN5v7JCdyqaGx-LWUBEn3TNfwzUgkKFBjVg40p2MPZVuzA@mail.gmail.com> <CAFO2ENm5f=ZBnU+x51y_DC_d=drPZXhgBn+YVMXWug=7y_BZXA@mail.gmail.com> <CAFO2ENk-zE5AzffCmu60=JLkj25Z08vbBLPNiaVEdq3sg88BLg@mail.gmail.com> <CAG2Zi21jHtjo1WaOD2Bbwd3JgaVg6RArEgY7QoYM1uM2Ou7hZw@mail.gmail.com> <CANzuiCg0WTShfVW_DHrmy7EehPFhFAvh8dHAGJjftiYnbfxSyw@mail.gmail.com> <CACsn0ck+jwwuushJsYDhxc9mwO36dyw0hrOm5VrDx=fw_9N7rg@mail.gmail.com> <CANzuiCjDBSaikQJJD39-EZpwSaedue7xq1dZiCJ3ciMruEbdhQ@mail.gmail.com> <CABN231rmmXmaSDxhXwX2TG+DVqDjHVWkoVVxUZsC2oQOv8X20g@mail.gmail.com> <CAG2Zi232a4RQp-bq=Tk6YptOLne-QPi6Osnsfg0j+AMku6skJQ@mail.gmail.com>
To: Christopher Patton <cpatton=40cloudflare.com@dmarc.ietf.org>
X-Mailer: Apple Mail (2.3693.40.0.1.81)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/wrQA-x8hQWHGg-zjFc9zaa2IriU>
Subject: Re: [Ppm] Uploading report shares directly to aggregators
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 25 Jan 2022 19:49:32 -0000

Sorry for chiming in late here. I agree that something like Option #2 is =
perhaps best, assuming we spell it in such a way that an ingestion =
service can be accommodated without protocol changes.=20

Best,
Chris

> On Jan 21, 2022, at 6:31 PM, Christopher Patton =
<cpatton=3D40cloudflare.com@dmarc.ietf.org> wrote:
>=20
> Thanks, Tim!
>=20
> =20
> A few of us were batting this around earlier and concluded that the =
problem of aggregators having disjoint views of what shares exist occurs =
no matter what: even in the single upload model, we would expect that =
aggregators will implement some kind of retention policy for submitted =
reports to stop their storage costs from growing indefinitely. So maybe =
a helper discards unaggregated shares after 14 days but its leader does =
so after 21. Even if they use the same retention policy, we can't =
guarantee that they will discard shares at the same moment. If a =
collector takes long enough to issue a `CollectReq` (in some VDAFs like =
`poplar1`, the aggregate phase cannot begin until the collector supplies =
the aggregation parameter, i.e., the candidate prefixes), then we are in =
the same error case that is introduced by partial success of a split =
upload. That means that the PPM protocol has to handle this error case =
no matter what (and we believe it already does).
>=20
> If I may try to summarize: the main downside of Option #2 is new =
failure modes, but similar failure modes already exist, and we'll have =
to deal with them in much the same way. I'm inclined to agree, although =
there is still a possibility of solving the existing problem in a =
different way. I'll note that this question is currently tracked by =
https://github.com/abetterinternet/ppm-specification/issues/180. That =
said, I'm inclined to agree with your analysis here.
>=20
> My preference would be to take Option #2.
>=20
> Hope y'all have a good weekend,
> Chris P.
>=20
>=20
> --=20
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm


From nobody Tue Jan 25 11:55:40 2022
Return-Path: <ekr@rtfm.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id BD84A3A0A1D for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 11:55:32 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.897
X-Spam-Level: 
X-Spam-Status: No, score=-1.897 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_NONE=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=rtfm-com.20210112.gappssmtp.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id htPfO--kz7u5 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 11:55:30 -0800 (PST)
Received: from mail-io1-xd29.google.com (mail-io1-xd29.google.com [IPv6:2607:f8b0:4864:20::d29]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 7D0C13A09BA for <ppm@ietf.org>; Tue, 25 Jan 2022 11:55:30 -0800 (PST)
Received: by mail-io1-xd29.google.com with SMTP id h23so24958555iol.11 for <ppm@ietf.org>; Tue, 25 Jan 2022 11:55:30 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=rtfm-com.20210112.gappssmtp.com; s=20210112; h=mime-version:from:date:message-id:subject:to; bh=WIcdyGHvLrDbWWU3aeBrnsDDTUBf8mMJdmXFP01DBMU=; b=zlBCbHUQmumoh1zlPPg/Nf6PAeqGAzZTr9blIBykR74ABYK9i4QFpbC4sROjV+ek/Q BI2KL7xkn15u2aeQaFc6e0LUW9gze+noXihw7Pyw5LNFdBWCOokd1ED0wHf8T13CFCv3 lhNctRBbmbe74C+qAYAsUqVNhgI/5XaMTtJ6T7uOMM6lIfKUMdp/DMu+qnxwJCb8miD8 qVcMbyaXEBRXSx1pAplWw7Om9/DrsQQFukXKLHgwrDk50syPhxkse4v4hO9hcktPKWyT Ji4yL22FureMiIZ5m7MtqjdnzXz1w0LWLc1/DajCO/c4BJCbI7vpQnUMvPxWiJ3cZoGw CKSw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:from:date:message-id:subject:to; bh=WIcdyGHvLrDbWWU3aeBrnsDDTUBf8mMJdmXFP01DBMU=; b=CGpXEF6vyNyEBfpneh2MZQ2yNOOqVLqheVnojZ7Sr4wSfJSryvL6RQfDBr/3FZ4/9J wCgl79EUhb4PeTu38fxeEPbV87ncNViIuemGr9Y1hErA5b46RST/ntTHuh5O+f59t1x+ m/xoF1YYsNBt1fUE6hf6gsEmu1UJh7VPd474dczL6k22L6EOC4Ini6231J5py4Vz1XhQ zejgd36+tuGNNH99KfkRSLZ80CxwGcSK9/BJ+h71Z82gsMMK5WI5N92CzPrz6/st87gE TxtzyDOYMI4WNQE2QkL09JeuwBqnkoSI8kabLCvd+MwAzYZmwmAFvB7LV/7xn3Qx+Dz6 l8xA==
X-Gm-Message-State: AOAM533qBFh2DEbpzeDyxOXD/BNe6iJgw/5FYeWq7rsUgNeGLuiROjZK hHZvjR8khki5P0GxYJ8TgIzWoYaI127JSG0qlM7cbhviYQ8nkg==
X-Google-Smtp-Source: ABdhPJy5OA8rv8i+JB0U28Myy1U19O6bzPCa7ujkdnXJ5tvOHDpmTwj628OpduBWMeHlLyO+RAivibcnEkdc3svOM30=
X-Received: by 2002:a5e:8504:: with SMTP id i4mr6716275ioj.149.1643140528084;  Tue, 25 Jan 2022 11:55:28 -0800 (PST)
MIME-Version: 1.0
From: Eric Rescorla <ekr@rtfm.com>
Date: Tue, 25 Jan 2022 11:54:52 -0800
Message-ID: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com>
To: ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000007e2f2f05d66d775d"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/WCvuU7zL-nHD6KBdVMhYHbcQ0OQ>
Subject: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 25 Jan 2022 19:55:38 -0000

--0000000000007e2f2f05d66d775d
Content-Type: text/plain; charset="UTF-8"

Hi folks,

I've reviewed this PR and while I think thre are good ideas, it's
also not ready.

  https://github.com/abetterinternet/ppm-specification/pull/174/

The current design has clients always uploading to the leader. I agree
that it's limiting, but I think it's also fully worked out.

This design replaces that by having the clients always uploading to
the aggregators, except maybe they actually go through some relatively
loosely specified "intermediary". This seems not to be fully worked
out. For instance:

- What is the behavior of that intermediary?
  - Does it batch?
  - Does it do some kind of anti-fraud stuff and if so does this
    get propagated to the helper?

- What is the threat model vis-a-vis the intermediary?

- How do we deal with cases where the aggregators have inconsistent
  data, e.g., because an upload fails? Does this just fail? Do you
  retry? How do you know? Do we need some protocol to find out
  in advance so you don't get errors?

I think we need to answer this kind of design question prior to making
this change.

-Ekr

--0000000000007e2f2f05d66d775d
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hi folks,<br><br>I&#39;ve reviewed this PR and while I thi=
nk thre are good ideas, it&#39;s<br>also not ready.<br><br>=C2=A0 <a href=
=3D"https://github.com/abetterinternet/ppm-specification/pull/174/">https:/=
/github.com/abetterinternet/ppm-specification/pull/174/</a><br><br>The curr=
ent design has clients always uploading to the leader. I agree<br>that it&#=
39;s limiting, but I think it&#39;s also fully worked out.<br><br>This desi=
gn replaces that by having the clients always uploading to<br>the aggregato=
rs, except maybe they actually go through some relatively<br>loosely specif=
ied &quot;intermediary&quot;. This seems not to be fully worked<br>out. For=
 instance:<br><br>- What is the behavior of that intermediary?<br>=C2=A0 - =
Does it batch?<br>=C2=A0 - Does it do some kind of anti-fraud stuff and if =
so does this<br>=C2=A0 =C2=A0 get propagated to the helper?<br><br>- What i=
s the threat model vis-a-vis the intermediary?<br><br>- How do we deal with=
 cases where the aggregators have inconsistent<br>=C2=A0 data, e.g., becaus=
e an upload fails? Does this just fail? Do you<br>=C2=A0 retry? How do you =
know? Do we need some protocol to find out<br>=C2=A0 in advance so you don&=
#39;t get errors?<br><br>I think we need to answer this kind of design ques=
tion prior to making<br>this change.<br><br>-Ekr<br><br>=C2=A0 <br><br><br>=
<br><br><br></div>

--0000000000007e2f2f05d66d775d--


From nobody Tue Jan 25 15:59:09 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 946D83A1720 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 15:59:00 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id xmWe8I_jdWA1 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 15:58:55 -0800 (PST)
Received: from mail-pj1-x102e.google.com (mail-pj1-x102e.google.com [IPv6:2607:f8b0:4864:20::102e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id DC3E73A1710 for <ppm@ietf.org>; Tue, 25 Jan 2022 15:58:55 -0800 (PST)
Received: by mail-pj1-x102e.google.com with SMTP id w12-20020a17090a528c00b001b276aa3aabso3225594pjh.0 for <ppm@ietf.org>; Tue, 25 Jan 2022 15:58:55 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=0tR9NuA8/J10cCw5MJfTR3Pfoy9dLC6TnegTt9qjAY4=; b=tK6cCAP4/t9M/++pG+ueYcRfmfZCKxo94u9GA5sZAjkf6Mq1WuL6gjGa8tsfbR7S7K EQWCPOGRKRS1gT/PlciTZLLcsSMIoqUFrebbY6cbj+6sCbkpgBDIZlOW0Y585+5/tWoQ 6hWJxLZd8Kvetfj9xPj55nXGLwYO6A5k1FFZU=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=0tR9NuA8/J10cCw5MJfTR3Pfoy9dLC6TnegTt9qjAY4=; b=c73mw8+vv7D9vE7lVQ+caouYUfsf/sxsUEgi1xdqzu6DWPBMocfV+sva3m1otZ4LT7 B82T0bQOZXkpSdf6t6RsagA1c3ukH0ADo9QD3yOHHm6crLpb+dFd+tuL50++HebkQ942 mB1PouiMKiKoahSwKu/DdqN7HdVFbtB+nyoRzRJDrN4qAnP4kqjYgmhd30X7j62QNfl/ OSzwRggN+HrRH43iWWYnAIbUX1NTBOXbjmIa3l8SNmfqYQfePSOETQMhYLMWkSJF0ioG Rq6RN18w/G8N6ANNzaxBjZPKaM7F0awRClINGxAUNLxJU+J+SGFvk1h78J+usZ0uJagf E1dQ==
X-Gm-Message-State: AOAM531K9RZfkrKxAr9WkNPmQpNvItRQ2IPEOMrdnpNNypFE2QZtWD6T dWUgvdLFJgKZU6bqWjydeo918nL46hyNwT9oT+m9xCDLATEySw==
X-Google-Smtp-Source: ABdhPJxPx737DPFX8JQBP8Xe3HW+evrSlEVDRVrrscKvRjjuujdJaUGqPdLDeAjimpsdg//oYF2BENSSRNOqHvHG+9E=
X-Received: by 2002:a17:903:31d5:b0:14b:53b7:cbd3 with SMTP id v21-20020a17090331d500b0014b53b7cbd3mr10658538ple.29.1643155134456; Tue, 25 Jan 2022 15:58:54 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com>
In-Reply-To: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Tue, 25 Jan 2022 15:58:43 -0800
Message-ID: <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com>
To: Eric Rescorla <ekr@rtfm.com>
Cc: ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000001a2a0b05d670dec7"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/ei1jUkCbTVrwUnaQ6IzuwUO0APU>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 25 Jan 2022 23:59:08 -0000

--0000000000001a2a0b05d670dec7
Content-Type: text/plain; charset="UTF-8"

Hi EKR,

At a high level, your questions seem to get at a more fundamental question
of the scope of the PPM spec over all: If the spec permits deployments with
an intermediary, does the spec need to fully specify how that intermediary
works? The consensus we reached in [3] seems to be that the answer is "no",
however we might have missed something important. See detailed responses to
your questions in-line.



> - What is the behavior of that intermediary?
>   - Does it batch?
>

Right now the PR presumes this behavior is out-of-scope. Can you think of a
reason why this behavior needs to be fleshed out in the core PPM spec?



>   - Does it do some kind of anti-fraud stuff and if so does this
>     get propagated to the helper?
>

By "anti-fraud" stuff I'm assuming you're referring to "Sybil attacks" as
described in [1]. Please correct me if I'm wrong.

Right now PPM does nothing about Sybil attacks, except that it provides an
extension point that a deployment could use for mitigating these. In this
regard, I don't think the situation has changed much.

   - Before #174, the leader controlled the set of reports that are
   aggregated, so it could mount a Sybil attack by simply stuffing the batch
   with reports it made up.
   - After #174, there may be an intermediary taht controls the set of
   reports that are aggregated, so it could mount a Sybil attack in exactly
   the same way.

Either way, an honest aggregator needs to do something to prevent Sybil
attacks. I don't think we've talked about this specifically in this WG yet,
but a good option might be the "Rate-Limited Tokens" spec being worked on
in the Privacy Pass WG [2]. A goal of ours should be to allow the
aggregators to deploy something like [2] *even when running with an
intermediary*. I believe this is possible: The clients still need to touch
the aggregator to get its HPKE public key, so it should be possible for an
aggregator to issue a token at that time.

Note that I have yet to take a close look at [2], so I might be completely
off base here :)



> - What is the threat model vis-a-vis the intermediary?
>

I don't believe that allowing an intermediary changes the threat model
beyond the fact that it gets to pick the set of reports that are
aggregated. As discussed above, the leader currently has the same power, so
this is already a problem. In fact, in the absence of a defense, any
aggregator can mount a Sybil attack by simply uploading report shares to
its peers. We don't want to have to trust any aggregator more than the
others, so a defense will be needed whether we take #174 or not.



> - How do we deal with cases where the aggregators have inconsistent
>   data, e.g., because an upload fails? Does this just fail? Do you
>   retry? How do you know? Do we need some protocol to find out
>   in advance so you don't get errors?
>

This was discussed in detail in the previous thread, so I'll just leave a
link to Tim's summary: [3]. The gist is that many of the new failure modes
that arise are similar to failure modes that already exist, and we'll need
to deal with these in much the same way. Thus dealing with split uploads
doesn't appear to require too much additional plumbing. If we missed
something in that discussion, please let us know!


Best,
Chris P.


[1] Issue discussing Sybil attacks:
https://github.com/abetterinternet/ppm-specification/issues/89
[2] Internet-Draft: Rate-Limited Token Issuance Protocol:
https://tfpauly.github.io/privacy-proxy/draft-privacypass-rate-limit-tokens.html
[3] Tim's summary about split-upload failure modes:
https://mailarchive.ietf.org/arch/msg/ppm/OSfhcossIrEPqPhIEW5CkVcJhik/

--0000000000001a2a0b05d670dec7
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div class=3D"gmail_quote">Hi EKR,</div><div class=3D"gmai=
l_quote"><br></div><div class=3D"gmail_quote">At a high level, your questio=
ns seem to get at a more fundamental question of the scope of the PPM spec =
over all: If the spec permits deployments with an intermediary, does the sp=
ec need to fully specify how that intermediary works? The consensus we reac=
hed in [3] seems to be that the answer is &quot;no&quot;, however we might =
have missed something important. See detailed responses to your questions i=
n-line.<br></div><div class=3D"gmail_quote"></div><div class=3D"gmail_quote=
"><br></div><div class=3D"gmail_quote"><div>=C2=A0</div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr">- What is the behavior of=
 that intermediary?<br>=C2=A0 - Does it batch?<br></div></blockquote><div><=
br></div><div>Right now the PR presumes this behavior is out-of-scope. Can =
you think of a reason why this behavior needs to be fleshed out in the core=
 PPM spec?<br></div><div><br></div><div>=C2=A0</div><blockquote class=3D"gm=
ail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,=
204,204);padding-left:1ex"><div dir=3D"ltr">=C2=A0 - Does it do some kind o=
f anti-fraud stuff and if so does this<br>=C2=A0 =C2=A0 get propagated to t=
he helper?<br></div></blockquote><div><br></div><div>By &quot;anti-fraud&qu=
ot; stuff I&#39;m assuming you&#39;re referring to &quot;Sybil attacks&quot=
; as described in [1]. Please correct me if I&#39;m wrong.<br></div><div><b=
r></div><div>Right now PPM does nothing about Sybil attacks, except that it=
 provides an extension point that a deployment could use for mitigating the=
se. In this regard, I don&#39;t think the situation has changed much.<br></=
div><div><ul><li>Before #174, the leader controlled the set of reports that=
 are aggregated, so it could mount a Sybil attack by simply stuffing the ba=
tch with reports it made up. <br></li><li>After #174, there may be an inter=
mediary taht controls the set of reports that are aggregated, so it could m=
ount a Sybil attack in exactly the same way.</li></ul><div>Either way, an h=
onest aggregator needs to do something to prevent Sybil attacks. I don&#39;=
t think we&#39;ve talked about this specifically in this WG yet, but a good=
 option might be the &quot;Rate-Limited Tokens&quot; spec being worked on i=
n the Privacy Pass WG [2]. A goal of ours should be to allow the aggregator=
s to deploy something like [2] *even when running with an intermediary*. I =
believe this is possible: The clients still need to touch the aggregator to=
 get its HPKE public key, so it should be possible for an aggregator to iss=
ue a token at that time.</div><div><br></div><div>Note that I have yet to t=
ake a close look at [2], so I might be completely off base here :)</div></d=
iv><div><br></div><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex"><div dir=3D"ltr">- What is the threat model vis-a-vis the interm=
ediary?<br></div></blockquote><div><br></div><div>I don&#39;t believe that =
allowing an intermediary changes the threat model beyond the fact that it g=
ets to pick the set of reports that are aggregated. As discussed above, the=
 leader currently has the same power, so this is already a problem. In fact=
, in the absence of a defense, any aggregator can mount a Sybil attack by s=
imply uploading report shares to its peers. We don&#39;t want to have to tr=
ust any aggregator more than the others, so a defense will be needed whethe=
r we take #174 or not.<br></div><div><br></div><div>=C2=A0</div><blockquote=
 class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px so=
lid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">- How do we deal wi=
th cases where the aggregators have inconsistent<br>=C2=A0 data, e.g., beca=
use an upload fails? Does this just fail? Do you<br>=C2=A0 retry? How do yo=
u know? Do we need some protocol to find out<br>=C2=A0 in advance so you do=
n&#39;t get errors?<br></div></blockquote><div><br></div><div>This was disc=
ussed in detail in the previous thread, so I&#39;ll just leave a link to Ti=
m&#39;s summary: [3]. The gist is that many of the new failure modes that a=
rise are similar to failure modes that already exist, and we&#39;ll need to=
 deal with these in much the same way. Thus dealing with split uploads does=
n&#39;t appear to require too much additional plumbing. If we missed someth=
ing in that discussion, please let us know!<br></div></div><div class=3D"gm=
ail_quote"><br></div><div class=3D"gmail_quote"><br></div><div class=3D"gma=
il_quote">Best,</div><div class=3D"gmail_quote">Chris P.</div><div class=3D=
"gmail_quote"><br></div><div class=3D"gmail_quote"><br></div><div class=3D"=
gmail_quote">[1] Issue discussing Sybil attacks: <a href=3D"https://github.=
com/abetterinternet/ppm-specification/issues/89">https://github.com/abetter=
internet/ppm-specification/issues/89</a></div><div class=3D"gmail_quote">[2=
] Internet-Draft: Rate-Limited Token Issuance Protocol: <a href=3D"https://=
tfpauly.github.io/privacy-proxy/draft-privacypass-rate-limit-tokens.html">h=
ttps://tfpauly.github.io/privacy-proxy/draft-privacypass-rate-limit-tokens.=
html</a></div><div class=3D"gmail_quote">[3] Tim&#39;s summary about split-=
upload failure modes: <a href=3D"https://mailarchive.ietf.org/arch/msg/ppm/=
OSfhcossIrEPqPhIEW5CkVcJhik/">https://mailarchive.ietf.org/arch/msg/ppm/OSf=
hcossIrEPqPhIEW5CkVcJhik/</a></div></div>

--0000000000001a2a0b05d670dec7--


From nobody Tue Jan 25 16:21:49 2022
Return-Path: <ekr@rtfm.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 8A0BA3A1810 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 16:21:47 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.897
X-Spam-Level: 
X-Spam-Status: No, score=-1.897 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_NONE=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=rtfm-com.20210112.gappssmtp.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 7YxUMK22s132 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 16:21:45 -0800 (PST)
Received: from mail-il1-x133.google.com (mail-il1-x133.google.com [IPv6:2607:f8b0:4864:20::133]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 776343A180E for <ppm@ietf.org>; Tue, 25 Jan 2022 16:21:45 -0800 (PST)
Received: by mail-il1-x133.google.com with SMTP id i1so12282501ils.5 for <ppm@ietf.org>; Tue, 25 Jan 2022 16:21:45 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=rtfm-com.20210112.gappssmtp.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=ecNCvrZaNlJQ+TrUyHHu6cmcpk8U6bk6SjkjkW+ip0I=; b=eJJaQ4NwvexVmStsV7VxL179PjpgPXFgpcq+Vn1CFUvtVYce6G5GYI/YwOdRMvYjuO ZwHCDIZG9/pXsYGW6nq7mKTrqlDr/Ru3KGiUEE16oP5zToVmBXcjStsIi1/CX21S8rHd 4BwQN+6HXClbL5cDs03K5AodP5McghhUWw8zoas5EgGVJNtYyWyT3WaAni4yX2c5KUjv LCpn3V42VUlppJcJJ/lhvfvfs9PAdvMU4W5FBJkgjWhwz7NFP7bWXIjYpLUCpy/DtM47 4tcOcY3V1ZOpFoIhGyXm6QlhncP2t5AkWgpp+kX+VfqCenBMQidcZYFWqB2TLSddMs/h yD4w==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=ecNCvrZaNlJQ+TrUyHHu6cmcpk8U6bk6SjkjkW+ip0I=; b=Z+T1UALHy4O3R1GIlKpoahXcM3QFuFl5klHQhoDK1ziahm0vdtPCaHQz8jdT3CZMA6 1gAMF+0gUI+Nb+koH+GEQHtcfrQZ+5mROIHYjzZbNsOEuMYQv+L5WY/yrIJJIU2VeJrD hRY5yeIE7SVlG+NVLLMWkqUK/1UjPfXCWuqrKXQA833T3kDQCe15Q8ub9dDUEBJ7GbGx OXLN3aLGKLVAbjCoc3zp9XeaAB+9nJ8ymnhWim44mqHrivVmz1FsjavN0cKh4COXVEz0 jV1HYt0ODrEQdoi1HvL+lZ65jmpkiXdNLOwctDVdaGQ8tdhFjbtk2VZv5qTIuawCM0rD +3ng==
X-Gm-Message-State: AOAM533j2+VH18D7BKH8jNY2RgOnRtXrBtDYvJxheLmOHc2F5LiVFtwo MbBQ7JJbeJF/E3OdOmQTq63Ie5qqD/M23qIZgcSdI9q/6mqggQ==
X-Google-Smtp-Source: ABdhPJzEck1XBxs/yEgIEt3iM8N9CT6widbHPSPHUNYeiG57UlY3iyPYRyeHruLsmFTcoWFK7jVnUnPCxESkNRNLUFI=
X-Received: by 2002:a05:6e02:1946:: with SMTP id x6mr13047293ilu.10.1643156503106;  Tue, 25 Jan 2022 16:21:43 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com>
In-Reply-To: <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com>
From: Eric Rescorla <ekr@rtfm.com>
Date: Tue, 25 Jan 2022 16:21:07 -0800
Message-ID: <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com>
To: Christopher Patton <cpatton@cloudflare.com>
Cc: ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000adac6b05d6712fa3"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/hGprb4GGNXxSwkFx9qpllUYJPko>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 26 Jan 2022 00:21:48 -0000

--000000000000adac6b05d6712fa3
Content-Type: text/plain; charset="UTF-8"

>>   - What is the behavior of that intermediary?
>>      - Does it batch?
>
> Right now the PR presumes this behavior is out-of-scope. Can you
> think of a reason why this behavior needs to be fleshed out in the
> core PPM spec?

Yes, because it goes to the responsibilities of the various players
in the system. For instance, if the intermediary batches, it may
be necessary for the other parts of the system to know the batch
size in order to reliably avoid prematurely aggregating time
windows.


>>    - Does it do some kind of anti-fraud stuff and if so does this
>>       get propagated to the helper?
>
> By "anti-fraud" stuff I'm assuming you're referring to "Sybil attacks"
> as described in [1]. Please correct me if I'm wrong.

Well, it's sort of unclear what "Sybil attacks" means in the context
of unauthenticated clients, but consider the case where you have
an attacker who runs a bot farm and is interested in changing the
output of your computation.



> Right now PPM does nothing about Sybil attacks, except that it
> provides an extension point that a deployment could use for
> mitigating these. In this regard, I don't think the situation has
> changed much.

> Before #174, the leader controlled the set of reports that are
> aggregated, so it could mount a Sybil attack by simply stuffing the
> batch with reports it made up.
> After #174, there may be an intermediary taht controls the set of
> reports that are aggregated, so it could mount a Sybil attack in
> exactly the same way.

This is my point about the threat model. There are (at least) two
threats:

1. Privacy threats where an attacker is trying to determine
   individual report values.
2. Integrity threats where an attacker is trying to affect
   the result of the measurement.

It's true that we need some mechanism to prevent Sybil attacks of the
former category that doesnt depend on trusting the leader, but the
second category of attack does not have this problem because the
leader can already affect the result of the computation, and so in the
current model, the leader can already take care of this, e.g., by IP
reputation, but but we have to ask if the intermediary will do so and
so now it needs to be trusted to produce a correct computation.
And of course there are also attacks where the intermediary
*removes* shares, which can't be protected by PrivacyPass and hte like.


> Either way, an honest aggregator needs to do something to prevent
> Sybil attacks. I don't think we've talked about this specifically in
> this WG yet, but a good option might be the "Rate-Limited Tokens" spec
> being worked on in the Privacy Pass WG [2]. A goal of ours should be
> to allow the aggregators to deploy something like [2] *even when
> running with an intermediary*. I believe this is possible: The clients
> still need to touch the aggregator to get its HPKE public key, so it
> should be possible for an aggregator to issue a token at that time.
>
> Note that I have yet to take a close look at [2], so I might be
> completely off base here :)

I don't think that addresses my concern, for the reasons above.



>>   - What is the threat model vis-a-vis the intermediary?
>
> I don't believe that allowing an intermediary changes the threat model
> beyond the fact that it gets to pick the set of reports that are
> aggregated. As discussed above, the leader currently has the same
> power, so this is already a problem. In fact, in the absence of a
> defense, any aggregator can mount a Sybil attack by simply uploading
> report shares to its peers. We don't want to have to trust any
> aggregator more than the others, so a defense will be needed whether
> we take #174 or not.

See above. We already trust the leader to help produce a correct
result, but this design *also* trusts the intermediary to do so,
so now it's yet another trusted part of the system.


>> - How do we deal with cases where the aggregators have inconsistent
>>   data, e.g., because an upload fails? Does this just fail? Do you
>>   retry? How do you know? Do we need some protocol to find out
>>   in advance so you don't get errors?
>
> This was discussed in detail in the previous thread, so I'll just
> leave a link to Tim's summary: [3]. The gist is that many of the new
> failure modes that arise are similar to failure modes that already
> exist, and we'll need to deal with these in much the same way. Thus
> dealing with split uploads doesn't appear to require too much
> additional plumbing. If we missed something in that discussion, please
> let us know!

Yeah, I don't think this conclusion is correct. It's correct that
all of these failure modes exist to some extent with the current
model, but scale matters. In particular, with some reasonable
design choices (for instance, having the same retention policy
on each helper, to use Tim's example), you can reduce the number
of inconsistencies to a fairly small number when you can just report
it as error. However, it's quite plausible that in the environment
contemplated by #174 where each client uploads separately, you
see regular inconsistencies at a reasonable fraction (single digit
percentages?) of the number of reports, in which case you can't
just write them off as measurement error but actually need a system
to recover from them. This design doesn't provide such a recoevery
mechanism, AFAICT.

-Ekr

--000000000000adac6b05d6712fa3
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">&gt;&gt; =C2=A0 - What is the behavior of that intermediar=
y?<br>&gt;&gt; =C2=A0 =C2=A0 =C2=A0- Does it batch?<br>&gt;<br>&gt; Right n=
ow the PR presumes this behavior is out-of-scope. Can you<br>&gt; think of =
a reason why this behavior needs to be fleshed out in the<br>&gt; core PPM =
spec?<br><br>Yes, because it goes to the responsibilities of the various pl=
ayers<br>in the system. For instance, if the intermediary batches, it may<b=
r>be necessary for the other parts of the system to know the batch<br>size =
in order to reliably avoid prematurely aggregating time<br>windows.<br><br>=
<br>&gt;&gt; =C2=A0 =C2=A0- Does it do some kind of anti-fraud stuff and if=
 so does this<br>&gt;&gt; =C2=A0 =C2=A0 =C2=A0 get propagated to the helper=
?<br>&gt;<br>&gt; By &quot;anti-fraud&quot; stuff I&#39;m assuming you&#39;=
re referring to &quot;Sybil attacks&quot;<br>&gt; as described in [1]. Plea=
se correct me if I&#39;m wrong.<br><br>Well, it&#39;s sort of unclear what =
&quot;Sybil attacks&quot; means in the context<br>of unauthenticated client=
s, but consider the case where you have<br>an attacker who runs a bot farm =
and is interested in changing the<br>output of your computation.<br><br><br=
><br>&gt; Right now PPM does nothing about Sybil attacks, except that it<br=
>&gt; provides an extension point that a deployment could use for<br>&gt; m=
itigating these. In this regard, I don&#39;t think the situation has<br>&gt=
; changed much.<br><br>&gt; Before #174, the leader controlled the set of r=
eports that are<br>&gt; aggregated, so it could mount a Sybil attack by sim=
ply stuffing the<br>&gt; batch with reports it made up.<br>&gt; After #174,=
 there may be an intermediary taht controls the set of<br>&gt; reports that=
 are aggregated, so it could mount a Sybil attack in<br>&gt; exactly the sa=
me way.<br><br>This is my point about the threat model. There are (at least=
) two<br>threats:<br><br>1. Privacy threats where an attacker is trying to =
determine<br>=C2=A0 =C2=A0individual report values.<br>2. Integrity threats=
 where an attacker is trying to affect<br>=C2=A0 =C2=A0the result of the me=
asurement.<br><br>It&#39;s true that we need some mechanism to prevent Sybi=
l attacks of the<br>former category that doesnt depend on trusting the lead=
er, but the<br>second category of attack does not have this problem because=
 the<br>leader can already affect the result of the computation, and so in =
the<br>current model, the leader can already take care of this, e.g., by IP=
<br>reputation, but but we have to ask if the intermediary will do so and<b=
r>so now it needs to be trusted to produce a correct computation.<br>And of=
 course there are also attacks where the intermediary<br>*removes* shares, =
which can&#39;t be protected by PrivacyPass and hte like.<br><br><br>&gt; E=
ither way, an honest aggregator needs to do something to prevent<br>&gt; Sy=
bil attacks. I don&#39;t think we&#39;ve talked about this specifically in<=
br>&gt; this WG yet, but a good option might be the &quot;Rate-Limited Toke=
ns&quot; spec<br>&gt; being worked on in the Privacy Pass WG [2]. A goal of=
 ours should be<br>&gt; to allow the aggregators to deploy something like [=
2] *even when<br>&gt; running with an intermediary*. I believe this is poss=
ible: The clients<br>&gt; still need to touch the aggregator to get its HPK=
E public key, so it<br>&gt; should be possible for an aggregator to issue a=
 token at that time.<br>&gt; <br>&gt; Note that I have yet to take a close =
look at [2], so I might be<br>&gt; completely off base here :)<br><br>I don=
&#39;t think that addresses my concern, for the reasons above.<br><br>=C2=
=A0<br><br>&gt;&gt; =C2=A0 - What is the threat model vis-a-vis the interme=
diary?<br>&gt;<br>&gt; I don&#39;t believe that allowing an intermediary ch=
anges the threat model<br>&gt; beyond the fact that it gets to pick the set=
 of reports that are<br>&gt; aggregated. As discussed above, the leader cur=
rently has the same<br>&gt; power, so this is already a problem. In fact, i=
n the absence of a<br>&gt; defense, any aggregator can mount a Sybil attack=
 by simply uploading<br>&gt; report shares to its peers. We don&#39;t want =
to have to trust any<br>&gt; aggregator more than the others, so a defense =
will be needed whether<br>&gt; we take #174 or not.<br><br>See above. We al=
ready trust the leader to help produce a correct<br>result, but this design=
 *also* trusts the intermediary to do so,<br>so now it&#39;s yet another tr=
usted part of the system.<br>=C2=A0<br><br>&gt;&gt; - How do we deal with c=
ases where the aggregators have inconsistent<br>&gt;&gt; =C2=A0 data, e.g.,=
 because an upload fails? Does this just fail? Do you<br>&gt;&gt; =C2=A0 re=
try? How do you know? Do we need some protocol to find out<br>&gt;&gt; =C2=
=A0 in advance so you don&#39;t get errors?<br>&gt;<br>&gt; This was discus=
sed in detail in the previous thread, so I&#39;ll just<br>&gt; leave a link=
 to Tim&#39;s summary: [3]. The gist is that many of the new<br>&gt; failur=
e modes that arise are similar to failure modes that already<br>&gt; exist,=
 and we&#39;ll need to deal with these in much the same way. Thus<br>&gt; d=
ealing with split uploads doesn&#39;t appear to require too much<br>&gt; ad=
ditional plumbing. If we missed something in that discussion, please<br>&gt=
; let us know!<br><br>Yeah, I don&#39;t think this conclusion is correct. I=
t&#39;s correct that<br>all of these failure modes exist to some extent wit=
h the current<br>model, but scale matters. In particular, with some reasona=
ble<br>design choices (for instance, having the same retention policy<br>on=
 each helper, to use Tim&#39;s example), you can reduce the number<br>of in=
consistencies to a fairly small number when you can just report<br>it as er=
ror. However, it&#39;s quite plausible that in the environment<br>contempla=
ted by #174 where each client uploads separately, you<br>see regular incons=
istencies at a reasonable fraction (single digit<br>percentages?) of the nu=
mber of reports, in which case you can&#39;t<br>just write them off as meas=
urement error but actually need a system<br>to recover from them. This desi=
gn doesn&#39;t provide such a recoevery<br>mechanism, AFAICT.<br><br>-Ekr<b=
r><br><br><br><br></div>

--000000000000adac6b05d6712fa3--


From nobody Tue Jan 25 17:01:38 2022
Return-Path: <caw@heapingbits.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 352493A1976 for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 17:01:37 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.098
X-Spam-Level: 
X-Spam-Status: No, score=-2.098 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=heapingbits.net header.b=fEiS6oK4; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=DPkvAtbm
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id P3EkqZ1WioeR for <ppm@ietfa.amsl.com>; Tue, 25 Jan 2022 17:01:32 -0800 (PST)
Received: from out4-smtp.messagingengine.com (out4-smtp.messagingengine.com [66.111.4.28]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 4D8023A196D for <ppm@ietf.org>; Tue, 25 Jan 2022 17:01:32 -0800 (PST)
Received: from compute1.internal (compute1.nyi.internal [10.202.2.41]) by mailout.nyi.internal (Postfix) with ESMTP id 5FB4D5C0271; Tue, 25 Jan 2022 20:01:31 -0500 (EST)
Received: from mailfrontend1 ([10.202.2.162]) by compute1.internal (MEProxy); Tue, 25 Jan 2022 20:01:31 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=heapingbits.net; h=cc:cc:content-transfer-encoding:content-type:date:date:from :from:in-reply-to:in-reply-to:message-id:mime-version:references :reply-to:sender:subject:subject:to:to; s=fm3; bh=0d6ZTqDaq9l+FJ 8S28d9BBuE7TJXLGAbTvRTTXH7LD0=; b=fEiS6oK4Q5U1ANLf0EbeRJjJZtNmD3 Af2aSiyti1Kz4xf1lQb9JqjtTdniWAcE9hgK2magvm3coHXEu30huOtvmfdxZZdd d1PN7eCkgW1dAnfVOH/4yBxxu0F6yQuZpGgiAtYBbcNhIhwpy44+fEff3wTaZGgz q2RPFUc6FMTGHm0pjaYRTb043XSs5ktaOipEjU25mIMnyY9TsZIcAaDlicaAqVAD CI+QbVQXiqSRKUEEo4k2NKZvBZPhUYyLmEp22vjqfoGtWV1ol+epdN/QdasAgagD PG2nntiQhbAl1b0zbvZX+5EmHjTe3Cun5/4IcUjtVpVCoIruO6J7iX7g==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:cc:content-transfer-encoding :content-type:date:date:from:from:in-reply-to:in-reply-to :message-id:mime-version:references:reply-to:sender:subject :subject:to:to:x-me-proxy:x-me-proxy:x-me-sender:x-me-sender :x-sasl-enc; s=fm1; bh=0d6ZTqDaq9l+FJ8S28d9BBuE7TJXLGAbTvRTTXH7L D0=; b=DPkvAtbmQtN3/x4K9LKUgguT20EY3reR6H3HEQLTWY0SaM4rkvk0w74UZ x9j1zywHFktiZ/0u3KBKYVUIXEWZKCnmB1J8vGdH5qj/b3SL8RRnZvziGPyckPCD r7CL+eapmggt+t3LZEEHRT4BRETQbR/LB3GmdrrNHThOlG9InhuLHEsGIlMd3A0X iEkfLHdsl5lqh/p42JnETPbfV+DCeB8FSKw+LLJPSH8R/qdNASP4y7BrmegjMHso M2fUsssEhir3uyV/SIPdBLEK7RCTzZuI44KHpRMqT4+Nvd/b/X7k2wdC21tLKamy TCPOwTHxPDmrKl+IFs63azT09PEZg==
X-ME-Sender: <xms:ap3wYfQX99ac9FuzCMYhph0EWyE1AB6XQ1pUvFG5LUJbG_VLsjfFRg> <xme:ap3wYQyHoHsirxpGx2MvKHcVpwz74A1RQTZl3bV75b6wZeB_7oYKOmND8biUOgBW- kgV_yMJT46mesPsoiw>
X-ME-Received: <xmr:ap3wYU2gHruebjtddDr5ESxtK0g1ksWN0hn3QOSwLAiHmu_-cEWqZawWBGca-jp922Q8eGwdW8ibbUzx45ngaKHB-JmucMgseA>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrfedtgdeftdcutefuodetggdotefrodftvf curfhrohhfihhlvgemucfhrghsthforghilhdpqfgfvfdpuffrtefokffrpgfnqfghnecu uegrihhlohhuthemuceftddtnecusecvtfgvtghiphhivghnthhsucdlqddutddtmdenuc fjughrpegtggfuhfgjfffgkfhfvffosehtqhhmtdhhtdejnecuhfhrohhmpeevhhhrihhs thhophhhvghrucghohhougcuoegtrgifsehhvggrphhinhhgsghithhsrdhnvghtqeenuc ggtffrrghtthgvrhhnpeeuheekledtfeetffethedvhffffedvfeduueeiueejvefhvdeg ieetffdtvdffjeenucffohhmrghinhepihgvthhfrdhorhhgnecuvehluhhsthgvrhfuih iivgeptdenucfrrghrrghmpehmrghilhhfrhhomheptggrfieshhgvrghpihhnghgsihht shdrnhgvth
X-ME-Proxy: <xmx:a53wYfDa7gEK4KIBTSyhW7Y2fo1HgIlI1iQrOpENLhWKVDGHDmI2KA> <xmx:a53wYYic6pzZ0I40n84V9fuh7tX4SpInF4HJkC6Cq9pw4woFKIKdhg> <xmx:a53wYTpyzVYrkoNg5kPPGJGB-lS6DtZ7xZvnZdPbaSqsJAwwF2x91w> <xmx:a53wYUZUzGkKDXoVwyDH_eXEuOvpW0Ty1qW9yKDQ0OgGEF_P12BWNA>
Received: by mail.messagingengine.com (Postfix) with ESMTPA; Tue, 25 Jan 2022 20:01:30 -0500 (EST)
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 15.0 \(3693.40.0.1.81\))
From: Christopher Wood <caw@heapingbits.net>
In-Reply-To: <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com>
Date: Tue, 25 Jan 2022 17:01:28 -0800
Cc: Christopher Patton <cpatton@cloudflare.com>, ppm@ietf.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net>
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com>
To: Eric Rescorla <ekr@rtfm.com>
X-Mailer: Apple Mail (2.3693.40.0.1.81)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/tEpt6bYZ5A6euAncWhHvfBzOeDY>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 26 Jan 2022 01:01:37 -0000

> On Jan 25, 2022, at 4:21 PM, Eric Rescorla <ekr@rtfm.com> wrote:
>=20
> >>   - What is the behavior of that intermediary?
> >>      - Does it batch?
> >
> > Right now the PR presumes this behavior is out-of-scope. Can you
> > think of a reason why this behavior needs to be fleshed out in the
> > core PPM spec?
>=20
> Yes, because it goes to the responsibilities of the various players
> in the system. For instance, if the intermediary batches, it may
> be necessary for the other parts of the system to know the batch
> size in order to reliably avoid prematurely aggregating time
> windows.

I don=E2=80=99t think this needs to be detailed in the spec. At the end =
of the day, I see no technical reason why the aggregators =E2=80=94 =
leader and helper =E2=80=94 should distinguish one entity sending many =
reports from many entities sending few reports. And if they=E2=80=99re =
not distinguishing in this manner, reports can be batched and sent =
however makes sense for a given task configuration. Certainly the =
batcher needs to know about the batch parameters. So the answer here =
seems to be: don=E2=80=99t use a batcher unless that batcher knows what =
it=E2=80=99s doing.

> >>    - Does it do some kind of anti-fraud stuff and if so does this
> >>       get propagated to the helper?
> >
> > By "anti-fraud" stuff I'm assuming you're referring to "Sybil =
attacks"
> > as described in [1]. Please correct me if I'm wrong.
>=20
> Well, it's sort of unclear what "Sybil attacks" means in the context
> of unauthenticated clients, but consider the case where you have
> an attacker who runs a bot farm and is interested in changing the
> output of your computation.

I think anti-fraud stuff should remain out of scope for the core spec.

> > Right now PPM does nothing about Sybil attacks, except that it
> > provides an extension point that a deployment could use for
> > mitigating these. In this regard, I don't think the situation has
> > changed much.
>=20
> > Before #174, the leader controlled the set of reports that are
> > aggregated, so it could mount a Sybil attack by simply stuffing the
> > batch with reports it made up.
> > After #174, there may be an intermediary taht controls the set of
> > reports that are aggregated, so it could mount a Sybil attack in
> > exactly the same way.
>=20
> This is my point about the threat model. There are (at least) two
> threats:
>=20
> 1. Privacy threats where an attacker is trying to determine
>    individual report values.
> 2. Integrity threats where an attacker is trying to affect
>    the result of the measurement.
>=20
> It's true that we need some mechanism to prevent Sybil attacks of the
> former category that doesnt depend on trusting the leader, but the
> second category of attack does not have this problem because the
> leader can already affect the result of the computation, and so in the
> current model, the leader can already take care of this, e.g., by IP
> reputation, but but we have to ask if the intermediary will do so and
> so now it needs to be trusted to produce a correct computation.
> And of course there are also attacks where the intermediary
> *removes* shares, which can't be protected by PrivacyPass and hte =
like.

Hmm, I think this is a deployment specific problem, rather than a =
problem that the core spec needs to address. In particular, I don=E2=80=99=
t see a difference between a malicious ingester dropping select reports, =
and an adversary prohibiting specific clients from uploading shares.=20

>=20
> > Either way, an honest aggregator needs to do something to prevent
> > Sybil attacks. I don't think we've talked about this specifically in
> > this WG yet, but a good option might be the "Rate-Limited Tokens" =
spec
> > being worked on in the Privacy Pass WG [2]. A goal of ours should be
> > to allow the aggregators to deploy something like [2] *even when
> > running with an intermediary*. I believe this is possible: The =
clients
> > still need to touch the aggregator to get its HPKE public key, so it
> > should be possible for an aggregator to issue a token at that time.
> >=20
> > Note that I have yet to take a close look at [2], so I might be
> > completely off base here :)
>=20
> I don't think that addresses my concern, for the reasons above.
>=20
> =20
>=20
> >>   - What is the threat model vis-a-vis the intermediary?
> >
> > I don't believe that allowing an intermediary changes the threat =
model
> > beyond the fact that it gets to pick the set of reports that are
> > aggregated. As discussed above, the leader currently has the same
> > power, so this is already a problem. In fact, in the absence of a
> > defense, any aggregator can mount a Sybil attack by simply uploading
> > report shares to its peers. We don't want to have to trust any
> > aggregator more than the others, so a defense will be needed whether
> > we take #174 or not.
>=20
> See above. We already trust the leader to help produce a correct
> result, but this design *also* trusts the intermediary to do so,
> so now it's yet another trusted part of the system.

I don=E2=80=99t really see this as a difference. The leader was already =
doing two functions: ingesting reports and aggregating. This proposal =
allows these functions to be split up without substantially altering the =
trust model.

> =20
>=20
> >> - How do we deal with cases where the aggregators have inconsistent
> >>   data, e.g., because an upload fails? Does this just fail? Do you
> >>   retry? How do you know? Do we need some protocol to find out
> >>   in advance so you don't get errors?
> >
> > This was discussed in detail in the previous thread, so I'll just
> > leave a link to Tim's summary: [3]. The gist is that many of the new
> > failure modes that arise are similar to failure modes that already
> > exist, and we'll need to deal with these in much the same way. Thus
> > dealing with split uploads doesn't appear to require too much
> > additional plumbing. If we missed something in that discussion, =
please
> > let us know!
>=20
> Yeah, I don't think this conclusion is correct. It's correct that
> all of these failure modes exist to some extent with the current
> model, but scale matters. In particular, with some reasonable
> design choices (for instance, having the same retention policy
> on each helper, to use Tim's example), you can reduce the number
> of inconsistencies to a fairly small number when you can just report
> it as error. However, it's quite plausible that in the environment
> contemplated by #174 where each client uploads separately, you
> see regular inconsistencies at a reasonable fraction (single digit
> percentages?) of the number of reports, in which case you can't
> just write them off as measurement error but actually need a system
> to recover from them. This design doesn't provide such a recoevery
> mechanism, AFAICT.

Well, it _could_ bake in such recovery mechanisms, right? (I agree that =
more work is needed on the PR, but this doesn=E2=80=99t seem like a =
dealbreaker to me.)

Best,
Chris

>=20
> -Ekr
>=20
>=20
>=20
>=20
> --=20
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm


From nobody Wed Jan 26 08:01:54 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 97BCE3A15BE for <ppm@ietfa.amsl.com>; Wed, 26 Jan 2022 08:01:52 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.099
X-Spam-Level: 
X-Spam-Status: No, score=-2.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id NMKfPAlqaq5D for <ppm@ietfa.amsl.com>; Wed, 26 Jan 2022 08:01:48 -0800 (PST)
Received: from mail-pg1-x536.google.com (mail-pg1-x536.google.com [IPv6:2607:f8b0:4864:20::536]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 31B4D3A15BD for <ppm@ietf.org>; Wed, 26 Jan 2022 08:01:47 -0800 (PST)
Received: by mail-pg1-x536.google.com with SMTP id g2so21484950pgo.9 for <ppm@ietf.org>; Wed, 26 Jan 2022 08:01:47 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=LE4ebxfng0jOhB4GNENXfSI/6hbysPSTgN2K8b/5phY=; b=jDlPuka1ot5R2/6qcAkKipy6wfumXyaGDDa1DDH4eybRQgU2lZm16iGi+Lz+B33b9O U8yXYSdloUAdDi8XnhwmtV92EccPnq/g9QaTOpP7FqgRIQhmE8my7NQM2NmQdRJkiDVk H9pn6bC2eSLE6CuG03mwg8Yfcvdl3Akig2biE=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=LE4ebxfng0jOhB4GNENXfSI/6hbysPSTgN2K8b/5phY=; b=L+WD5E5X7XBUc3D2yXnRMW9GUyUpwiNwiqaVi1tiaYvZ3XcqzC3r/GPE5PpyBisIg6 l1qKDKv8hhcDTQFpJruFygiqUWBAG3xo/VEokiwHcK2XKpK7KAN8NXT9C1W67DuZP4Fm epTm//dSQYdesoAPixU8uNsQidIzLMGj8YxscxxOkm9iWJWZ/k88tf6jSVMXoDpJMKx2 ukgjViUyfVEKtsZGYYcQOo4NbHoByEObUOlt7vJYFiKjuKfuDL85i3Kfwh8gJwy0+svc wWkv2ggU/BX39EDwSZxqjX4CMJ8HWOwZRbFWzxjs+DqsBoRUaex5mwT9wj1E+mcjdfNQ RXTA==
X-Gm-Message-State: AOAM531FbOcBZ6hS4i1n7/GdqEIRbkS5ZULQYQhQXrtqPHo63j7wCXYN +iPRO7ZdOH4F7cn/jDPsbN+IB22TQVxVh8rTB6sVaLZtYEP8Ew==
X-Google-Smtp-Source: ABdhPJztbs5JyV3hdsecQWdP+8cCY3gJwE4LbQpbCpRGOd9Y9pI7wvyWsvTnY91lVin3XJsLanxgvUiV4nSAeP0pymI=
X-Received: by 2002:a63:cf13:: with SMTP id j19mr18924794pgg.279.1643212906627;  Wed, 26 Jan 2022 08:01:46 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com> <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net>
In-Reply-To: <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Wed, 26 Jan 2022 08:01:35 -0800
Message-ID: <CAG2Zi22+sv1L_omrWqHeDo2JqVr=XESUAUix2pJEr1WqKUjHRQ@mail.gmail.com>
To: Christopher Wood <caw@heapingbits.net>
Cc: Eric Rescorla <ekr@rtfm.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="000000000000977f0a05d67e5192"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/SiVeLEL7l9YfGpI3XKbGokmSOn0>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 26 Jan 2022 16:01:53 -0000

--000000000000977f0a05d67e5192
Content-Type: text/plain; charset="UTF-8"

I agree that, in deployments where clients upload report shares to
aggregators directly, data loss is more likely because the upload operation
is no longer atomic. One way we know how to mitigate this is to put an
ingestor in front of the aggregators, which would be functionally
equivalent to the current leader-upload architecture. I think all we need
to do is figure out how to spell this in the PR. In particular, it should
be possible for a deployment to basically work exactly like the
leader-upload architecture.

--000000000000977f0a05d67e5192
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div><span class=3D"gmail-im">I agree that, in deployments=
 where clients upload report shares to aggregators directly, data loss is m=
ore likely because the upload operation is no longer atomic. One way we kno=
w how to mitigate this is to put an ingestor in front of the aggregators, w=
hich would be functionally equivalent to the current leader-upload architec=
ture. I think all we need to do is figure out how to spell this in the PR. =
In particular, it should be possible for a deployment to basically work exa=
ctly like the leader-upload architecture.<br></span></div></div>

--000000000000977f0a05d67e5192--


From nobody Wed Jan 26 12:36:47 2022
Return-Path: <timg@letsencrypt.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 6B4C43A1F9A for <ppm@ietfa.amsl.com>; Wed, 26 Jan 2022 12:36:45 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.674
X-Spam-Level: 
X-Spam-Status: No, score=-2.674 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.576, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=letsencrypt.org
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ZrDFQR-MSy0N for <ppm@ietfa.amsl.com>; Wed, 26 Jan 2022 12:36:40 -0800 (PST)
Received: from mail-qt1-x82e.google.com (mail-qt1-x82e.google.com [IPv6:2607:f8b0:4864:20::82e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id C3BB13A1F96 for <ppm@ietf.org>; Wed, 26 Jan 2022 12:36:40 -0800 (PST)
Received: by mail-qt1-x82e.google.com with SMTP id y8so735521qtn.8 for <ppm@ietf.org>; Wed, 26 Jan 2022 12:36:40 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=letsencrypt.org; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=8o8iW8DFjUFGA+DBG8KObk5BxBYp87+lxaF1+X9KDMI=; b=fyJDN06xjlhESCeYwCNsoZFlkko9C3cYi5EN64PuBASKl/oU5x8kMgX1YuBz+Jf5oT cbNHIzDGyNsh7K8wLSe8s7REx3aDBnxhAuUvoVPpUhu8ofpWsC70eMkIXuzGM3/454iO CAqvKfati6kfjHIuJwQy7KH2tAJgA9jwflBbE=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=8o8iW8DFjUFGA+DBG8KObk5BxBYp87+lxaF1+X9KDMI=; b=0GwRqWu8+qS5Hwl6TX1vPsGhxA/J9eV2T35ncUUZh2e8+/zPA92Xd7PvAB74/UPmVT Hl7q9LIc2krEJAfxWewG0B9nfv3Pd/DTIx/OjiWCMRha5ynkucSrbXyRiUxY0UXZLa+0 QajpMmm26w4XgDhRD1gG3JljJ4RXUGbCOvm7gcOHtTBMjbD+0q6EDg7xjqw8HkCmToxE F8/CGzbCghS1PlGROoIBLUCznINcm3RZl0Y6/0oAAxRDaMhM/R2eUWGrt0eUgx6/8CCP IM/yKailZlk1rXZyse5WLOLAv8V63atFZOqWGw1EkOUkRIpNHGebeZRB/O4jF1WFNJa4 fQ6Q==
X-Gm-Message-State: AOAM530KstJGhbDoU06kfBHbfL6xEtXrmEU/LLIOw3BMpm+F4A7lYvCM /pZNWdN4djdbHawOje88UQpJGTndPgAhT3jYd5PMXw==
X-Google-Smtp-Source: ABdhPJzVQtJS2K6f2Bg9jQ6MWlw33hRuM3MSsjztgchIi/Odh50YkzBhtlMoGH2CHpMOEuvQkdUuPznyByc5EVG8GrU=
X-Received: by 2002:ac8:5c95:: with SMTP id r21mr356378qta.27.1643229399008; Wed, 26 Jan 2022 12:36:39 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com> <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net> <CAG2Zi22+sv1L_omrWqHeDo2JqVr=XESUAUix2pJEr1WqKUjHRQ@mail.gmail.com>
In-Reply-To: <CAG2Zi22+sv1L_omrWqHeDo2JqVr=XESUAUix2pJEr1WqKUjHRQ@mail.gmail.com>
From: Tim Geoghegan <timg@letsencrypt.org>
Date: Wed, 26 Jan 2022 12:36:28 -0800
Message-ID: <CABN231rTUXbxhGuUs0DsOdoY4uTyzaAKxROgEPANMHeiULpJ6g@mail.gmail.com>
To: Christopher Patton <cpatton=40cloudflare.com@dmarc.ietf.org>
Cc: Christopher Wood <caw@heapingbits.net>, Eric Rescorla <ekr@rtfm.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000009ce16605d68228eb"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/mJE2AUOfsbhQ09ZccnGwvplHhcI>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 26 Jan 2022 20:36:46 -0000

--0000000000009ce16605d68228eb
Content-Type: text/plain; charset="UTF-8"

I agree with Eric that we owe more text and diagrams illustrating how
deployments might use an ingestor, and to explain whether they affect the
threat model. My feeling/hope is that if we construct secure channels from
clients to the aggregators through any intermediaries, then the ingestor is
no more powerful than any network adversary who can drop messages.
Naturally my feelings aren't good enough for an IETF spec so we owe more
analysis and prescriptions for deployments.

However I'd like to know what we mean by "recovery mechanism". So the
scenario we are concerned about is one where each aggregator holds a
different subset of the reports generated by clients, such that the
intersection of the sets of reports held by leader and helper is (let's
say) only 80% of the generated reports. What does it mean to recover from
this? Does it mean that the system should still successfully aggregate over
the 80%, or that we should somehow augment the upload portion of the
protocol so that clients can retry and we get closer to 100%?

Tim

On Wed, Jan 26, 2022 at 8:01 AM Christopher Patton <cpatton=
40cloudflare.com@dmarc.ietf.org> wrote:

> I agree that, in deployments where clients upload report shares to
> aggregators directly, data loss is more likely because the upload operation
> is no longer atomic. One way we know how to mitigate this is to put an
> ingestor in front of the aggregators, which would be functionally
> equivalent to the current leader-upload architecture. I think all we need
> to do is figure out how to spell this in the PR. In particular, it should
> be possible for a deployment to basically work exactly like the
> leader-upload architecture.
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--0000000000009ce16605d68228eb
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>I agree with Eric that we owe more text and diagrams =
illustrating how deployments might use an ingestor, and to explain whether =
they affect the threat model. My feeling/hope is that if we construct secur=
e channels from clients to the aggregators through any intermediaries, then=
 the ingestor is no more powerful than any network adversary who can drop m=
essages. Naturally my feelings aren&#39;t good enough for an IETF spec so w=
e owe more analysis and prescriptions for deployments.</div><div><br></div>=
<div>However I&#39;d like to know what we mean by &quot;recovery mechanism&=
quot;. So the scenario we are concerned about is one where each aggregator =
holds a different subset of the reports generated by clients, such that the=
 intersection of the sets of reports held by leader and helper is (let&#39;=
s say) only 80% of the generated reports. What does it mean to recover from=
 this? Does it mean that the system should still successfully aggregate ove=
r the 80%, or that we should somehow augment the upload portion of the prot=
ocol so that clients can retry and we get closer to 100%?<br></div><div><br=
></div><div>Tim<br></div></div><br><div class=3D"gmail_quote"><div dir=3D"l=
tr" class=3D"gmail_attr">On Wed, Jan 26, 2022 at 8:01 AM Christopher Patton=
 &lt;cpatton=3D<a href=3D"mailto:40cloudflare.com@dmarc.ietf.org">40cloudfl=
are.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_q=
uote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,2=
04);padding-left:1ex"><div dir=3D"ltr"><div><span>I agree that, in deployme=
nts where clients upload report shares to aggregators directly, data loss i=
s more likely because the upload operation is no longer atomic. One way we =
know how to mitigate this is to put an ingestor in front of the aggregators=
, which would be functionally equivalent to the current leader-upload archi=
tecture. I think all we need to do is figure out how to spell this in the P=
R. In particular, it should be possible for a deployment to basically work =
exactly like the leader-upload architecture.<br></span></div></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--0000000000009ce16605d68228eb--


From nobody Thu Jan 27 05:59:08 2022
Return-Path: <ekr@rtfm.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 5C1B33A0AA2 for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 05:59:06 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.896
X-Spam-Level: 
X-Spam-Status: No, score=-6.896 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_NONE=0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=rtfm-com.20210112.gappssmtp.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 5fciTp3j0ucq for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 05:59:02 -0800 (PST)
Received: from mail-io1-xd2f.google.com (mail-io1-xd2f.google.com [IPv6:2607:f8b0:4864:20::d2f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 1BB073A0A9A for <ppm@ietf.org>; Thu, 27 Jan 2022 05:59:02 -0800 (PST)
Received: by mail-io1-xd2f.google.com with SMTP id p63so2762939iod.11 for <ppm@ietf.org>; Thu, 27 Jan 2022 05:59:02 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=rtfm-com.20210112.gappssmtp.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=lfTZURU9qat1AucRU4kCH2LctpIZE71bFWUmFzfk01E=; b=2o2+pClX+7cmLVQZXh0NR75ZRe6xH0EnKFIvI6z/U7uGIOUwt4N0pNj0XLzbJq4UZH 7kaDlG1bR6CsdYFGqw/lXiPFLNnhW9ODNW9DKKZ/znWCRuO9hx5qeCnsMKKAJTQ7cMv4 JlUMqcEWXc2rGIbPkhMbg9jz7u1vcdThzpkQ515wTtnxqvjRKrBW+HETJAzvONBgaaLH xGmLYYVw6Gja/HJMk+w7X6mAZdq/Q3n+dkW2sQ2g6lZE32R9NJj+16zMFLhjzqRwmZIx 0RtRqA0yoErxmTZN71GuLe0G4N8xsipJANI0+kvm/vS8lDcDJOGA7KFJ6FSL41EF1N5D dfKg==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=lfTZURU9qat1AucRU4kCH2LctpIZE71bFWUmFzfk01E=; b=58giprpyjJgo2r60fyAm2WVLuX5co8zltdR0O3KbHqS1fnO+vchpEfazV9WM/RIoXJ yg9rSFpoDHbgjlDT5VcZnH/uMHLz9u/z+vxi0DW72WtbcE+2c8kuFlmGPKQweUNEoxtQ jIERDI5Z8irJCR8vUuCO53mIaeE9END5NjC8ntNM/44t6vS+relxfbQHkv1mL8ToLizY a7fkL7Gq/YE2wTZhIYDOllWrwzDLB41GFB8NFS5VZoOmKfDzgAgp4hjuu+0Yc2uM2gsH 08S6M/foZaQkXMPrDg1KvpQUsLTwSVye5yw7G7JHHApv1f6GMvBieP8JgThtcIllW/3H hY/w==
X-Gm-Message-State: AOAM530bPk1e4O2qiAzbd0Ua7GPx5oVuwNRXP7k+faNP/DNdqivbZJSI vk0ImvyaTt2wD/SxA/5AOYOnBxeZSx4HDVE7FbLy9+NlXn8=
X-Google-Smtp-Source: ABdhPJx8tO5CUyFAcBVSWDWbWm//5qDU/maH6EIcnKYmZ7LfFjMy6fhIuJbIKgkuAL3mb0HRnjTsmWvfiC4rwB5zph4=
X-Received: by 2002:a05:6602:483:: with SMTP id y3mr2087407iov.0.1643291940448;  Thu, 27 Jan 2022 05:59:00 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com> <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net> <CAG2Zi22+sv1L_omrWqHeDo2JqVr=XESUAUix2pJEr1WqKUjHRQ@mail.gmail.com> <CABN231rTUXbxhGuUs0DsOdoY4uTyzaAKxROgEPANMHeiULpJ6g@mail.gmail.com>
In-Reply-To: <CABN231rTUXbxhGuUs0DsOdoY4uTyzaAKxROgEPANMHeiULpJ6g@mail.gmail.com>
From: Eric Rescorla <ekr@rtfm.com>
Date: Thu, 27 Jan 2022 05:58:24 -0800
Message-ID: <CABcZeBPSfjNpKD_CSmmDJyRWJyUs3dbP9tefEDOEiYmGynWnDA@mail.gmail.com>
To: Tim Geoghegan <timg@letsencrypt.org>
Cc: Christopher Patton <cpatton=40cloudflare.com@dmarc.ietf.org>,  Christopher Wood <caw@heapingbits.net>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="0000000000005f810c05d690b865"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/C9sNEODC-IcG0D785Kzjesualz4>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 27 Jan 2022 13:59:06 -0000

--0000000000005f810c05d690b865
Content-Type: text/plain; charset="UTF-8"

On Wed, Jan 26, 2022 at 12:36 PM Tim Geoghegan <timg@letsencrypt.org> wrote:

> I agree with Eric that we owe more text and diagrams illustrating how
> deployments might use an ingestor, and to explain whether they affect the
> threat model. My feeling/hope is that if we construct secure channels from
> clients to the aggregators through any intermediaries, then the ingestor is
> no more powerful than any network adversary who can drop messages
>

It's probably not, but that's actually quite a bad outcome because that
adversary is extremely powerful for this kind of measurement and surveying
application.

Perhaps we haven't done as great a job defining the threat model as one
would like, but ISTM that for unauthenticated clients, we're going to be
relying for accuracy fairly heavily on topological type stuff (e.g., client
IP address) in order to prevent naive stuffing of the ingestion channel.
But those defenses will not work against the ingestor.



> . Naturally my feelings aren't good enough for an IETF spec so we owe more
> analysis and prescriptions for deployments.
>
> However I'd like to know what we mean by "recovery mechanism". So the
> scenario we are concerned about is one where each aggregator holds a
> different subset of the reports generated by clients, such that the
> intersection of the sets of reports held by leader and helper is (let's
> say) only 80% of the generated reports. What does it mean to recover from
> this? Does it mean that the system should still successfully aggregate over
> the 80%, or that we should somehow augment the upload portion of the
> protocol so that clients can retry and we get closer to 100%?
>

Stepping back a level, a measurement protocol in which you only get 80% of
the data based on some unknown network conditions is likely to exhibit bias
in the measurements. So even with a conventional system you'd hope the
clients retry. But with a conventional system, if clients retry the data
eventually trickles in and your statistics correct themselves. But with
PPM, because we want to (1) avoid repeated aggregation of the same values
and (2) want to avoid aggregating small sets, it's a much bigger problem if
you just recompute the metric over the same value. So, that means that we
don't want to end up in a setting where (say) A and B have collectively
received 99% of the reports but actually have each only received 90% and
only have 90% overlap and then that 10% is lost to use forever.

-Ekr


>
> Tim
>
> On Wed, Jan 26, 2022 at 8:01 AM Christopher Patton <cpatton=
> 40cloudflare.com@dmarc.ietf.org> wrote:
>
>> I agree that, in deployments where clients upload report shares to
>> aggregators directly, data loss is more likely because the upload operation
>> is no longer atomic. One way we know how to mitigate this is to put an
>> ingestor in front of the aggregators, which would be functionally
>> equivalent to the current leader-upload architecture. I think all we need
>> to do is figure out how to spell this in the PR. In particular, it should
>> be possible for a deployment to basically work exactly like the
>> leader-upload architecture.
>> --
>> Ppm mailing list
>> Ppm@ietf.org
>> https://www.ietf.org/mailman/listinfo/ppm
>>
>

--0000000000005f810c05d690b865
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Wed, Jan 26, 2022 at 12:36 PM Tim =
Geoghegan &lt;<a href=3D"mailto:timg@letsencrypt.org">timg@letsencrypt.org<=
/a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0=
px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><=
div dir=3D"ltr"><div>I agree with Eric that we owe more text and diagrams i=
llustrating how deployments might use an ingestor, and to explain whether t=
hey affect the threat model. My feeling/hope is that if we construct secure=
 channels from clients to the aggregators through any intermediaries, then =
the ingestor is no more powerful than any network adversary who can drop me=
ssages</div></div></blockquote><div><br></div><div>It&#39;s probably not, b=
ut that&#39;s actually quite a bad outcome because that adversary is extrem=
ely powerful for this kind of measurement and surveying application.</div><=
div><br></div><div>Perhaps we haven&#39;t done as great a job defining the =
threat model as one would like, but ISTM that for unauthenticated clients, =
we&#39;re going to be relying for accuracy fairly heavily on topological ty=
pe stuff (e.g., client IP address) in order to prevent naive stuffing of th=
e ingestion channel. But those defenses will not work against the ingestor.=
<br></div><div><br></div><div>=C2=A0<br></div><blockquote class=3D"gmail_qu=
ote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,20=
4);padding-left:1ex"><div dir=3D"ltr"><div>. Naturally my feelings aren&#39=
;t good enough for an IETF spec so we owe more analysis and prescriptions f=
or deployments.</div><div><br></div><div>However I&#39;d like to know what =
we mean by &quot;recovery mechanism&quot;. So the scenario we are concerned=
 about is one where each aggregator holds a different subset of the reports=
 generated by clients, such that the intersection of the sets of reports he=
ld by leader and helper is (let&#39;s say) only 80% of the generated report=
s. What does it mean to recover from this? Does it mean that the system sho=
uld still successfully aggregate over the 80%, or that we should somehow au=
gment the upload portion of the protocol so that clients can retry and we g=
et closer to 100%?<br></div></div></blockquote><div><br></div><div>Stepping=
 back a level, a measurement protocol in which you only get 80% of the data=
 based on some unknown network conditions is likely to exhibit bias in the =
measurements. So even with a conventional system you&#39;d hope the clients=
 retry. But with a conventional system, if clients retry the data eventuall=
y trickles in and your statistics correct themselves. But with PPM, because=
 we want to (1) avoid repeated aggregation of the same values and (2) want =
to avoid aggregating small sets, it&#39;s a much bigger problem if you just=
 recompute the metric over the same value. So, that means that we don&#39;t=
 want to end up in a setting where (say) A and B have collectively received=
 99% of the reports but actually have each only received 90% and only have =
90% overlap and then that 10% is lost to use forever.<br></div><div><br></d=
iv><div>-Ekr</div><div>=C2=A0<br></div><blockquote class=3D"gmail_quote" st=
yle=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padd=
ing-left:1ex"><div dir=3D"ltr"><div></div><div><br></div><div>Tim<br></div>=
</div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">=
On Wed, Jan 26, 2022 at 8:01 AM Christopher Patton &lt;cpatton=3D<a href=3D=
"mailto:40cloudflare.com@dmarc.ietf.org" target=3D"_blank">40cloudflare.com=
@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" s=
tyle=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);pad=
ding-left:1ex"><div dir=3D"ltr"><div><span>I agree that, in deployments whe=
re clients upload report shares to aggregators directly, data loss is more =
likely because the upload operation is no longer atomic. One way we know ho=
w to mitigate this is to put an ingestor in front of the aggregators, which=
 would be functionally equivalent to the current leader-upload architecture=
. I think all we need to do is figure out how to spell this in the PR. In p=
articular, it should be possible for a deployment to basically work exactly=
 like the leader-upload architecture.<br></span></div></div>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>
</blockquote></div></div>

--0000000000005f810c05d690b865--


From nobody Thu Jan 27 06:16:07 2022
Return-Path: <ekr@rtfm.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 060173A0D2E for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 06:16:06 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.896
X-Spam-Level: 
X-Spam-Status: No, score=-6.896 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_NONE=0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=rtfm-com.20210112.gappssmtp.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 2sVfPY6iWkOa for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 06:16:00 -0800 (PST)
Received: from mail-io1-xd34.google.com (mail-io1-xd34.google.com [IPv6:2607:f8b0:4864:20::d34]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id A2AF73A0D2A for <ppm@ietf.org>; Thu, 27 Jan 2022 06:16:00 -0800 (PST)
Received: by mail-io1-xd34.google.com with SMTP id q204so3690386iod.8 for <ppm@ietf.org>; Thu, 27 Jan 2022 06:16:00 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=rtfm-com.20210112.gappssmtp.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=b+ZfzIzamJI9POzxcuVwCgA6QSgP499ehTJ9RsVRBYo=; b=EAuhQ7kt748oPSKFtDY6Qs8JfHuvANHxCsOUKiuB7ocGMFKlbp7Gg2UF/pXKcWwiiy BL+W+21ZMHDTxyrFb69+bGTSJESKH/jN3MBEiX+Nly6FqErQt/sKZt2wEZ22rhFVBIM0 VLcaDuoxUzzPeQlHNBRrSxrhcHrOjgfNEeoxZAru57dwHCxkkbN9OmLcCM1lkXpFLiS8 2e0XKMk7c+FBPMv7YcYuowkJYx87y5GmmVxle5dTCIOiaG5IwDkd3+piwTiBybzbUJVR +LLDeejLmcNziGaIgbq2HIc3xmfhqWQgdx0XkwH85LWreVWlrghgRkB2tHxM/vxfIV22 V4Jg==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=b+ZfzIzamJI9POzxcuVwCgA6QSgP499ehTJ9RsVRBYo=; b=h4eRNVNLN7A9z09+f2xACK2R0YYuNNR0AI8y4LP0bNiyYj8i18zgFkq/BBER3rFa7O 7LxCfGxDLnX4s7bQIUOHIHqltOYBr6ar7wUNPj+o8xjkNKFrD49T842kzHqpld+sms5e z61sGhoWZia1I5jkBllzjUlKkAy5T+tI+IHWRTuVG+bPjNX/oMaSSEqUuIytE4+VXiDT zTOWIZbVnt0OP+xkJeDJb6ER0OLhNsks7N4B/aC5pp+hFG9lkR+Vek/31gBNJk44e75M QQEfMwtKde3CAcIssdpIFD0zpUnVtAODus+GNYh3pdKmrAcV8A2yE/qeI7lzgbE6uiB5 XOqA==
X-Gm-Message-State: AOAM532aWRV196stzUDI9mZjHbbZJn+A87eiPD1lkbYG6qjCPgVqZ2fV ot5afDkG7gLlIOIgUzguRBK7f1keuqvRAdB8AjYfoph7uOc=
X-Google-Smtp-Source: ABdhPJwd3ER5XBHoyCRx24qk1L90pgOV0j2wR2BbUqwfRJ49y+Hu5jwR6egzCCroEfIxBgYbLotBhAqep8fem7aAHrM=
X-Received: by 2002:a6b:f80e:: with SMTP id o14mr2094808ioh.213.1643292959032;  Thu, 27 Jan 2022 06:15:59 -0800 (PST)
MIME-Version: 1.0
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com> <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net>
In-Reply-To: <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net>
From: Eric Rescorla <ekr@rtfm.com>
Date: Thu, 27 Jan 2022 06:15:23 -0800
Message-ID: <CABcZeBOjriSqG=8yGz5ibgeRL1uzJYn4n94seN8CdwQi4VS7HA@mail.gmail.com>
To: Christopher Wood <caw@heapingbits.net>
Cc: Christopher Patton <cpatton@cloudflare.com>, ppm@ietf.org
Content-Type: multipart/alternative; boundary="00000000000015de5105d690f504"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/dy638X3WKhYejrclJUbPRpMs9qk>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 27 Jan 2022 14:16:06 -0000

--00000000000015de5105d690f504
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, Jan 25, 2022 at 5:01 PM Christopher Wood <caw@heapingbits.net>
wrote:

>
> > On Jan 25, 2022, at 4:21 PM, Eric Rescorla <ekr@rtfm.com> wrote:
> >
> > >>   - What is the behavior of that intermediary?
> > >>      - Does it batch?
> > >
> > > Right now the PR presumes this behavior is out-of-scope. Can you
> > > think of a reason why this behavior needs to be fleshed out in the
> > > core PPM spec?
> >
> > Yes, because it goes to the responsibilities of the various players
> > in the system. For instance, if the intermediary batches, it may
> > be necessary for the other parts of the system to know the batch
> > size in order to reliably avoid prematurely aggregating time
> > windows.
>
> I don=E2=80=99t think this needs to be detailed in the spec. At the end o=
f the
> day, I see no technical reason why the aggregators =E2=80=94 leader and h=
elper =E2=80=94
> should distinguish one entity sending many reports from many entities
> sending few reports. And if they=E2=80=99re not distinguishing in this ma=
nner,
> reports can be batched and sent however makes sense for a given task
> configuration. Certainly the batcher needs to know about the batch
> parameters. So the answer here seems to be: don=E2=80=99t use a batcher u=
nless that
> batcher knows what it=E2=80=99s doing.
>

It's not a matter of knowing what you are doing but of coordination.

Suppose (for instance) that the ingestor batches on 10 day boundaries and
the minimum batch size for measurement is 7 days. How does the leader know
on day 16 that it can't aggregate period 7-14?


> > Right now PPM does nothing about Sybil attacks, except that it
> > > provides an extension point that a deployment could use for
> > > mitigating these. In this regard, I don't think the situation has
> > > changed much.
> >
> > > Before #174, the leader controlled the set of reports that are
> > > aggregated, so it could mount a Sybil attack by simply stuffing the
> > > batch with reports it made up.
> > > After #174, there may be an intermediary taht controls the set of
> > > reports that are aggregated, so it could mount a Sybil attack in
> > > exactly the same way.
> >
> > This is my point about the threat model. There are (at least) two
> > threats:
> >
> > 1. Privacy threats where an attacker is trying to determine
> >    individual report values.
> > 2. Integrity threats where an attacker is trying to affect
> >    the result of the measurement.
> >
> > It's true that we need some mechanism to prevent Sybil attacks of the
> > former category that doesnt depend on trusting the leader, but the
> > second category of attack does not have this problem because the
> > leader can already affect the result of the computation, and so in the
> > current model, the leader can already take care of this, e.g., by IP
> > reputation, but but we have to ask if the intermediary will do so and
> > so now it needs to be trusted to produce a correct computation.
> > And of course there are also attacks where the intermediary
> > *removes* shares, which can't be protected by PrivacyPass and hte like.
>
> Hmm, I think this is a deployment specific problem, rather than a problem
> that the core spec needs to address. In particular, I don=E2=80=99t see a
> difference between a malicious ingester dropping select reports, and an
> adversary prohibiting specific clients from uploading shares.
>

See my response to Tim. In general, this system is not going to provide
correct results if an attacker completely controls communications between
all clients and the servers. However, note that the intermediary actually
is strictly more powerful than such an attacker, because it gets to see the
HTTP traffic not just the TLS traffic. For instance, suppose it uses
User-Agent to drop every submission from a Mac?



> > > Either way, an honest aggregator needs to do something to prevent
> > > Sybil attacks. I don't think we've talked about this specifically in
> > > this WG yet, but a good option might be the "Rate-Limited Tokens" spe=
c
> > > being worked on in the Privacy Pass WG [2]. A goal of ours should be
> > > to allow the aggregators to deploy something like [2] *even when
> > > running with an intermediary*. I believe this is possible: The client=
s
> > > still need to touch the aggregator to get its HPKE public key, so it
> > > should be possible for an aggregator to issue a token at that time.
> > >
> > > Note that I have yet to take a close look at [2], so I might be
> > > completely off base here :)
> >
> > I don't think that addresses my concern, for the reasons above.
> >
> >
> >
> > >>   - What is the threat model vis-a-vis the intermediary?
> > >
> > > I don't believe that allowing an intermediary changes the threat mode=
l
> > > beyond the fact that it gets to pick the set of reports that are
> > > aggregated. As discussed above, the leader currently has the same
> > > power, so this is already a problem. In fact, in the absence of a
> > > defense, any aggregator can mount a Sybil attack by simply uploading
> > > report shares to its peers. We don't want to have to trust any
> > > aggregator more than the others, so a defense will be needed whether
> > > we take #174 or not.
> >
> > See above. We already trust the leader to help produce a correct
> > result, but this design *also* trusts the intermediary to do so,
> > so now it's yet another trusted part of the system.
>
> I don=E2=80=99t really see this as a difference. The leader was already d=
oing two
> functions: ingesting reports and aggregating. This proposal allows these
> functions to be split up without substantially altering the trust model.
>

But this is precisely my point about the threat model: that the ingestor is
not some kind of untrusted service, it is an integral part of the system,
and has to be trusted, at least for the purpose of producing accurate
results. So, now we need an analysis of what the intermediary can or cannot
do.


> >> - How do we deal with cases where the aggregators have inconsistent
> > >>   data, e.g., because an upload fails? Does this just fail? Do you
> > >>   retry? How do you know? Do we need some protocol to find out
> > >>   in advance so you don't get errors?
> > >
> > > This was discussed in detail in the previous thread, so I'll just
> > > leave a link to Tim's summary: [3]. The gist is that many of the new
> > > failure modes that arise are similar to failure modes that already
> > > exist, and we'll need to deal with these in much the same way. Thus
> > > dealing with split uploads doesn't appear to require too much
> > > additional plumbing. If we missed something in that discussion, pleas=
e
> > > let us know!
> >
> > Yeah, I don't think this conclusion is correct. It's correct that
> > all of these failure modes exist to some extent with the current
> > model, but scale matters. In particular, with some reasonable
> > design choices (for instance, having the same retention policy
> > on each helper, to use Tim's example), you can reduce the number
> > of inconsistencies to a fairly small number when you can just report
> > it as error. However, it's quite plausible that in the environment
> > contemplated by #174 where each client uploads separately, you
> > see regular inconsistencies at a reasonable fraction (single digit
> > percentages?) of the number of reports, in which case you can't
> > just write them off as measurement error but actually need a system
> > to recover from them. This design doesn't provide such a recoevery
> > mechanism, AFAICT.
>
> Well, it _could_ bake in such recovery mechanisms, right? (I agree that
> more work is needed on the PR, but this doesn=E2=80=99t seem like a dealb=
reaker to
> me.)
>

Those recovery mechanisms may well be quite complicated.  Given that the
nominal purpose of this change is to reduce certain kinds of operational
complexity, I'd like to understand what the resulting additional complexity
of recovery is before making it.

-Ekr





> Best,
> Chris
>
> >
> > -Ekr
> >
> >
> >
> >
> > --
> > Ppm mailing list
> > Ppm@ietf.org
> > https://www.ietf.org/mailman/listinfo/ppm
>
>

--00000000000015de5105d690f504
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Tue, Jan 25, 2022 at 5:01 PM Chris=
topher Wood &lt;<a href=3D"mailto:caw@heapingbits.net" target=3D"_blank">ca=
w@heapingbits.net</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote"=
 style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);p=
adding-left:1ex"><br>
&gt; On Jan 25, 2022, at 4:21 PM, Eric Rescorla &lt;<a href=3D"mailto:ekr@r=
tfm.com" target=3D"_blank">ekr@rtfm.com</a>&gt; wrote:<br>
&gt; <br>
&gt; &gt;&gt;=C2=A0 =C2=A0- What is the behavior of that intermediary?<br>
&gt; &gt;&gt;=C2=A0 =C2=A0 =C2=A0 - Does it batch?<br>
&gt; &gt;<br>
&gt; &gt; Right now the PR presumes this behavior is out-of-scope. Can you<=
br>
&gt; &gt; think of a reason why this behavior needs to be fleshed out in th=
e<br>
&gt; &gt; core PPM spec?<br>
&gt; <br>
&gt; Yes, because it goes to the responsibilities of the various players<br=
>
&gt; in the system. For instance, if the intermediary batches, it may<br>
&gt; be necessary for the other parts of the system to know the batch<br>
&gt; size in order to reliably avoid prematurely aggregating time<br>
&gt; windows.<br>
<br>
I don=E2=80=99t think this needs to be detailed in the spec. At the end of =
the day, I see no technical reason why the aggregators =E2=80=94 leader and=
 helper =E2=80=94 should distinguish one entity sending many reports from m=
any entities sending few reports. And if they=E2=80=99re not distinguishing=
 in this manner, reports can be batched and sent however makes sense for a =
given task configuration. Certainly the batcher needs to know about the bat=
ch parameters. So the answer here seems to be: don=E2=80=99t use a batcher =
unless that batcher knows what it=E2=80=99s doing.<br></blockquote><br></di=
v><div class=3D"gmail_quote">It&#39;s not a matter of knowing what you are =
doing but of coordination.</div><div class=3D"gmail_quote"><br></div><div c=
lass=3D"gmail_quote">Suppose (for instance) that the ingestor batches on 10=
 day boundaries and the minimum batch size for measurement is 7 days. How d=
oes the leader know on day 16 that it can&#39;t aggregate period 7-14? <br>=
<div><br></div><br><blockquote class=3D"gmail_quote" style=3D"margin:0px 0p=
x 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
&gt; &gt; Right now PPM does nothing about Sybil attacks, except that it<br=
>
&gt; &gt; provides an extension point that a deployment could use for<br>
&gt; &gt; mitigating these. In this regard, I don&#39;t think the situation=
 has<br>
&gt; &gt; changed much.<br>
&gt; <br>
&gt; &gt; Before #174, the leader controlled the set of reports that are<br=
>
&gt; &gt; aggregated, so it could mount a Sybil attack by simply stuffing t=
he<br>
&gt; &gt; batch with reports it made up.<br>
&gt; &gt; After #174, there may be an intermediary taht controls the set of=
<br>
&gt; &gt; reports that are aggregated, so it could mount a Sybil attack in<=
br>
&gt; &gt; exactly the same way.<br>
&gt; <br>
&gt; This is my point about the threat model. There are (at least) two<br>
&gt; threats:<br>
&gt; <br>
&gt; 1. Privacy threats where an attacker is trying to determine<br>
&gt;=C2=A0 =C2=A0 individual report values.<br>
&gt; 2. Integrity threats where an attacker is trying to affect<br>
&gt;=C2=A0 =C2=A0 the result of the measurement.<br>
&gt; <br>
&gt; It&#39;s true that we need some mechanism to prevent Sybil attacks of =
the<br>
&gt; former category that doesnt depend on trusting the leader, but the<br>
&gt; second category of attack does not have this problem because the<br>
&gt; leader can already affect the result of the computation, and so in the=
<br>
&gt; current model, the leader can already take care of this, e.g., by IP<b=
r>
&gt; reputation, but but we have to ask if the intermediary will do so and<=
br>
&gt; so now it needs to be trusted to produce a correct computation.<br>
&gt; And of course there are also attacks where the intermediary<br>
&gt; *removes* shares, which can&#39;t be protected by PrivacyPass and hte =
like.<br>
<br>
Hmm, I think this is a deployment specific problem, rather than a problem t=
hat the core spec needs to address. In particular, I don=E2=80=99t see a di=
fference between a malicious ingester dropping select reports, and an adver=
sary prohibiting specific clients from uploading shares. <br></blockquote><=
div><br></div><div>See my response to Tim. In general, this system is not g=
oing to provide correct results if an attacker completely controls communic=
ations between all clients and the servers. However, note that the intermed=
iary actually is strictly more powerful than such an attacker, because it g=
ets to see the HTTP traffic not just the TLS traffic. For instance, suppose=
 it uses User-Agent to drop every submission from a Mac?<br></div><div><br>=
</div><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"margin:0p=
x 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
&gt; &gt; Either way, an honest aggregator needs to do something to prevent=
<br>
&gt; &gt; Sybil attacks. I don&#39;t think we&#39;ve talked about this spec=
ifically in<br>
&gt; &gt; this WG yet, but a good option might be the &quot;Rate-Limited To=
kens&quot; spec<br>
&gt; &gt; being worked on in the Privacy Pass WG [2]. A goal of ours should=
 be<br>
&gt; &gt; to allow the aggregators to deploy something like [2] *even when<=
br>
&gt; &gt; running with an intermediary*. I believe this is possible: The cl=
ients<br>
&gt; &gt; still need to touch the aggregator to get its HPKE public key, so=
 it<br>
&gt; &gt; should be possible for an aggregator to issue a token at that tim=
e.<br>
&gt; &gt; <br>
&gt; &gt; Note that I have yet to take a close look at [2], so I might be<b=
r>
&gt; &gt; completely off base here :)<br>
&gt; <br>
&gt; I don&#39;t think that addresses my concern, for the reasons above.<br=
>
&gt; <br>
&gt;=C2=A0 <br>
&gt; <br>
&gt; &gt;&gt;=C2=A0 =C2=A0- What is the threat model vis-a-vis the intermed=
iary?<br>
&gt; &gt;<br>
&gt; &gt; I don&#39;t believe that allowing an intermediary changes the thr=
eat model<br>
&gt; &gt; beyond the fact that it gets to pick the set of reports that are<=
br>
&gt; &gt; aggregated. As discussed above, the leader currently has the same=
<br>
&gt; &gt; power, so this is already a problem. In fact, in the absence of a=
<br>
&gt; &gt; defense, any aggregator can mount a Sybil attack by simply upload=
ing<br>
&gt; &gt; report shares to its peers. We don&#39;t want to have to trust an=
y<br>
&gt; &gt; aggregator more than the others, so a defense will be needed whet=
her<br>
&gt; &gt; we take #174 or not.<br>
&gt; <br>
&gt; See above. We already trust the leader to help produce a correct<br>
&gt; result, but this design *also* trusts the intermediary to do so,<br>
&gt; so now it&#39;s yet another trusted part of the system.<br>
<br>
I don=E2=80=99t really see this as a difference. The leader was already doi=
ng two functions: ingesting reports and aggregating. This proposal allows t=
hese functions to be split up without substantially altering the trust mode=
l.<br></blockquote><div><br></div><div>But this is precisely my point about=
 the threat model: that the ingestor is not some kind of untrusted service,=
 it is an integral part of the system, and has to be trusted, at least for =
the purpose of producing accurate results. So, now we need an analysis of w=
hat the intermediary can or cannot do. <br></div><div><br></div><div><br></=
div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex">
&gt; &gt;&gt; - How do we deal with cases where the aggregators have incons=
istent<br>
&gt; &gt;&gt;=C2=A0 =C2=A0data, e.g., because an upload fails? Does this ju=
st fail? Do you<br>
&gt; &gt;&gt;=C2=A0 =C2=A0retry? How do you know? Do we need some protocol =
to find out<br>
&gt; &gt;&gt;=C2=A0 =C2=A0in advance so you don&#39;t get errors?<br>
&gt; &gt;<br>
&gt; &gt; This was discussed in detail in the previous thread, so I&#39;ll =
just<br>
&gt; &gt; leave a link to Tim&#39;s summary: [3]. The gist is that many of =
the new<br>
&gt; &gt; failure modes that arise are similar to failure modes that alread=
y<br>
&gt; &gt; exist, and we&#39;ll need to deal with these in much the same way=
. Thus<br>
&gt; &gt; dealing with split uploads doesn&#39;t appear to require too much=
<br>
&gt; &gt; additional plumbing. If we missed something in that discussion, p=
lease<br>
&gt; &gt; let us know!<br>
&gt; <br>
&gt; Yeah, I don&#39;t think this conclusion is correct. It&#39;s correct t=
hat<br>
&gt; all of these failure modes exist to some extent with the current<br>
&gt; model, but scale matters. In particular, with some reasonable<br>
&gt; design choices (for instance, having the same retention policy<br>
&gt; on each helper, to use Tim&#39;s example), you can reduce the number<b=
r>
&gt; of inconsistencies to a fairly small number when you can just report<b=
r>
&gt; it as error. However, it&#39;s quite plausible that in the environment=
<br>
&gt; contemplated by #174 where each client uploads separately, you<br>
&gt; see regular inconsistencies at a reasonable fraction (single digit<br>
&gt; percentages?) of the number of reports, in which case you can&#39;t<br=
>
&gt; just write them off as measurement error but actually need a system<br=
>
&gt; to recover from them. This design doesn&#39;t provide such a recoevery=
<br>
&gt; mechanism, AFAICT.<br>
<br>
Well, it _could_ bake in such recovery mechanisms, right? (I agree that mor=
e work is needed on the PR, but this doesn=E2=80=99t seem like a dealbreake=
r to me.)<br></blockquote><div><br></div><div>Those recovery mechanisms may=
 well be quite complicated.=C2=A0 Given that the nominal purpose of this ch=
ange is to reduce certain kinds of operational complexity, I&#39;d like to =
understand what the resulting additional complexity of recovery is before m=
aking it.</div><div><br></div><div>-Ekr</div><div>=C2=A0<br></div><div><br>=
</div><div><br></div><div><br></div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex">
<br>
Best,<br>
Chris<br>
<br>
&gt; <br>
&gt; -Ekr<br>
&gt; <br>
&gt; <br>
&gt; <br>
&gt; <br>
&gt; -- <br>
&gt; Ppm mailing list<br>
&gt; <a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
&gt; <a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferre=
r" target=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
<br>
</blockquote></div></div>

--00000000000015de5105d690f504--


From nobody Thu Jan 27 14:00:21 2022
Return-Path: <caw@heapingbits.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 874943A12FA for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 14:00:19 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.098
X-Spam-Level: 
X-Spam-Status: No, score=-7.098 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_HI=-5, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=heapingbits.net header.b=b+JBsBgW; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=mQEbFTQn
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id dRv4cply_Iuu for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 14:00:14 -0800 (PST)
Received: from wout5-smtp.messagingengine.com (wout5-smtp.messagingengine.com [64.147.123.21]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 7B5253A12F6 for <ppm@ietf.org>; Thu, 27 Jan 2022 14:00:14 -0800 (PST)
Received: from compute5.internal (compute5.nyi.internal [10.202.2.45]) by mailout.west.internal (Postfix) with ESMTP id BB1433202349; Thu, 27 Jan 2022 17:00:13 -0500 (EST)
Received: from mailfrontend1 ([10.202.2.162]) by compute5.internal (MEProxy); Thu, 27 Jan 2022 17:00:14 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=heapingbits.net; h=cc:cc:content-transfer-encoding:content-type:date:date:from :from:in-reply-to:in-reply-to:message-id:mime-version:references :reply-to:sender:subject:subject:to:to; s=fm3; bh=KA8l3cSVl7gYYb AZJTL+K1bXzagWpEEO4iOYYJm/7So=; b=b+JBsBgWahm0VmQfgzcO+FMyK9UwYt 02ESkYcdPgb9yJhEbre3DsQyDVfqovu8WNWrjjapeR1VCkrWB5MhV+j9Qn94PYPv gITs70luqDJNW/08kBibRySU1MMCB6imJaOciShVUQg/iNqB04DorUkYtjxZgSLl tccbx/7TmgI92vj86fFNKTZePP9YsAa3qzRGkRxjmLrjsM33A4Gppp+Q5EjhL2KE 8IbhhBCZS1FQBgquHMnuf30d0uyEdN6fvdrUdL/1CoM3tyl7bld7uQIagp3D0kYh RcyEG1DEBAV6awESNB9yqE5AZ1+/LB383Dw787wOcWU4gr4D3fsvpnag==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:cc:content-transfer-encoding :content-type:date:date:from:from:in-reply-to:in-reply-to :message-id:mime-version:references:reply-to:sender:subject :subject:to:to:x-me-proxy:x-me-proxy:x-me-sender:x-me-sender :x-sasl-enc; s=fm1; bh=KA8l3cSVl7gYYbAZJTL+K1bXzagWpEEO4iOYYJm/7 So=; b=mQEbFTQnZ8DcC8s7Nh/UuNVOD1ei9o4SBnpcqGw8WZ2ijoDOieW0k6w3d V8/Yc45mmFyQkl9t9KXI+i/+xS5c1BLA/In3BgfEtrGBNOtoBxynBjTQStOzsx2p sJaqYq5So/Qdnxylh/0ZxXewI8UYedSb7zQt0dM0qxPszvnAjoRYlkGe3O4kawx6 NUG9qQZkHv2eLgsD5QdqOeSvdenpU8srNhCnsZzBAAxT6HTHMtXN7gPhksyW/8rJ 6jSoomYlFaL53kLJrc3Eh7+R4f83iPvdnvbC0V8mszPIgHiTvcUzAppi5viEs7Go USPpGyL9YVy1CfPh54dBddUZHI/HA==
X-ME-Sender: <xms:7BXzYX7fttq27JrGqhVJwM-_ZAhJqf2Lz6gk9RQahjRPIAa_EYM9jg> <xme:7BXzYc54xgA6e1yR1z1jIh6g2Dic5G3Ul-wqDYlweEA-0NvBHpBMph1co2mb4B9Ts 9sIs2tRMZZIjr2Pcmg>
X-ME-Received: <xmr:7BXzYefbUp4d9RXmTk88M5nrHPxlv2zT-dMrTCsLeYKQzOFZBzMcJvMR2lFX8NL1zYOFBZxx4ZMTlESvDnK8K9zS56Cmq7aADw>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrfeefgdduheeiucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucesvcftvggtihhpihgvnhhtshculddquddttddmne cujfgurheptggguffhjgffgffkfhfvofesthhqmhdthhdtjeenucfhrhhomhepvehhrhhi shhtohhphhgvrhcuhghoohguuceotggrfieshhgvrghpihhnghgsihhtshdrnhgvtheqne cuggftrfgrthhtvghrnhepueehkeeltdefteffteehvdfhffefvdefudeuieeujeevhfdv geeitefftddvffejnecuffhomhgrihhnpehivghtfhdrohhrghenucevlhhushhtvghruf hiiigvpedtnecurfgrrhgrmhepmhgrihhlfhhrohhmpegtrgifsehhvggrphhinhhgsghi thhsrdhnvght
X-ME-Proxy: <xmx:7BXzYYKXsCrJX2jY5Sy9FvfU6Gl4tmJGoXmBSct8RfCZko-aazzpGg> <xmx:7BXzYbISa3n0_KaVM5FDlQfmhd9sAhll5tubxM0Wbi5CaXdAnZ-d1w> <xmx:7BXzYRx-Yf7wzQqokj_YXyMm2dKxGUhd5nmBqkwkLpjbi2_3y7jpDQ> <xmx:7RXzYRhR_u26fizeoZ5KjEIQfifUlZu0NYcdbA_bSXAKWJl59Ql0WA>
Received: by mail.messagingengine.com (Postfix) with ESMTPA; Thu, 27 Jan 2022 17:00:12 -0500 (EST)
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 15.0 \(3693.40.0.1.81\))
From: Christopher Wood <caw@heapingbits.net>
In-Reply-To: <CABcZeBOjriSqG=8yGz5ibgeRL1uzJYn4n94seN8CdwQi4VS7HA@mail.gmail.com>
Date: Thu, 27 Jan 2022 14:00:10 -0800
Cc: Christopher Patton <cpatton@cloudflare.com>, ppm@ietf.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <714E486D-0597-4D4F-849E-82F21A58EBA5@heapingbits.net>
References: <CABcZeBMNa+gT8SYXoShLC_F1aAYO-g1NZ0b9myPbeuUTOYCekQ@mail.gmail.com> <CAG2Zi22po5zsZMHU-Wy5YRDq3kSkEt5dBEPNDUOr1KQA9mDnLg@mail.gmail.com> <CABcZeBMq3ad0NQev=F1kKPvsg+f60qcBAPMXKLBE98qROC8y5Q@mail.gmail.com> <04D21F4D-0124-4006-80A2-F01A82C0F46E@heapingbits.net> <CABcZeBOjriSqG=8yGz5ibgeRL1uzJYn4n94seN8CdwQi4VS7HA@mail.gmail.com>
To: Eric Rescorla <ekr@rtfm.com>
X-Mailer: Apple Mail (2.3693.40.0.1.81)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/_hJmwYuQ9kAXFMopCpiBu5CV8F8>
Subject: Re: [Ppm] PR #174 -- Direct Upload
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 27 Jan 2022 22:00:20 -0000

> On Jan 27, 2022, at 6:15 AM, Eric Rescorla <ekr@rtfm.com> wrote:
>=20
>=20
>=20
> On Tue, Jan 25, 2022 at 5:01 PM Christopher Wood <caw@heapingbits.net> =
wrote:
>=20
> > On Jan 25, 2022, at 4:21 PM, Eric Rescorla <ekr@rtfm.com> wrote:
> >=20
> > >>   - What is the behavior of that intermediary?
> > >>      - Does it batch?
> > >
> > > Right now the PR presumes this behavior is out-of-scope. Can you
> > > think of a reason why this behavior needs to be fleshed out in the
> > > core PPM spec?
> >=20
> > Yes, because it goes to the responsibilities of the various players
> > in the system. For instance, if the intermediary batches, it may
> > be necessary for the other parts of the system to know the batch
> > size in order to reliably avoid prematurely aggregating time
> > windows.
>=20
> I don=E2=80=99t think this needs to be detailed in the spec. At the =
end of the day, I see no technical reason why the aggregators =E2=80=94 =
leader and helper =E2=80=94 should distinguish one entity sending many =
reports from many entities sending few reports. And if they=E2=80=99re =
not distinguishing in this manner, reports can be batched and sent =
however makes sense for a given task configuration. Certainly the =
batcher needs to know about the batch parameters. So the answer here =
seems to be: don=E2=80=99t use a batcher unless that batcher knows what =
it=E2=80=99s doing.
>=20
> It's not a matter of knowing what you are doing but of coordination.
>=20
> Suppose (for instance) that the ingestor batches on 10 day boundaries =
and the minimum batch size for measurement is 7 days. How does the =
leader know on day 16 that it can't aggregate period 7-14?

(replied below)

>=20
>=20
> > > Right now PPM does nothing about Sybil attacks, except that it
> > > provides an extension point that a deployment could use for
> > > mitigating these. In this regard, I don't think the situation has
> > > changed much.
> >=20
> > > Before #174, the leader controlled the set of reports that are
> > > aggregated, so it could mount a Sybil attack by simply stuffing =
the
> > > batch with reports it made up.
> > > After #174, there may be an intermediary taht controls the set of
> > > reports that are aggregated, so it could mount a Sybil attack in
> > > exactly the same way.
> >=20
> > This is my point about the threat model. There are (at least) two
> > threats:
> >=20
> > 1. Privacy threats where an attacker is trying to determine
> >    individual report values.
> > 2. Integrity threats where an attacker is trying to affect
> >    the result of the measurement.
> >=20
> > It's true that we need some mechanism to prevent Sybil attacks of =
the
> > former category that doesnt depend on trusting the leader, but the
> > second category of attack does not have this problem because the
> > leader can already affect the result of the computation, and so in =
the
> > current model, the leader can already take care of this, e.g., by IP
> > reputation, but but we have to ask if the intermediary will do so =
and
> > so now it needs to be trusted to produce a correct computation.
> > And of course there are also attacks where the intermediary
> > *removes* shares, which can't be protected by PrivacyPass and hte =
like.
>=20
> Hmm, I think this is a deployment specific problem, rather than a =
problem that the core spec needs to address. In particular, I don=E2=80=99=
t see a difference between a malicious ingester dropping select reports, =
and an adversary prohibiting specific clients from uploading shares.=20
>=20
> See my response to Tim. In general, this system is not going to =
provide correct results if an attacker completely controls =
communications between all clients and the servers. However, note that =
the intermediary actually is strictly more powerful than such an =
attacker, because it gets to see the HTTP traffic not just the TLS =
traffic. For instance, suppose it uses User-Agent to drop every =
submission from a Mac?

(replied below)

>=20
> =20
> > > Either way, an honest aggregator needs to do something to prevent
> > > Sybil attacks. I don't think we've talked about this specifically =
in
> > > this WG yet, but a good option might be the "Rate-Limited Tokens" =
spec
> > > being worked on in the Privacy Pass WG [2]. A goal of ours should =
be
> > > to allow the aggregators to deploy something like [2] *even when
> > > running with an intermediary*. I believe this is possible: The =
clients
> > > still need to touch the aggregator to get its HPKE public key, so =
it
> > > should be possible for an aggregator to issue a token at that =
time.
> > >=20
> > > Note that I have yet to take a close look at [2], so I might be
> > > completely off base here :)
> >=20
> > I don't think that addresses my concern, for the reasons above.
> >=20
> > =20
> >=20
> > >>   - What is the threat model vis-a-vis the intermediary?
> > >
> > > I don't believe that allowing an intermediary changes the threat =
model
> > > beyond the fact that it gets to pick the set of reports that are
> > > aggregated. As discussed above, the leader currently has the same
> > > power, so this is already a problem. In fact, in the absence of a
> > > defense, any aggregator can mount a Sybil attack by simply =
uploading
> > > report shares to its peers. We don't want to have to trust any
> > > aggregator more than the others, so a defense will be needed =
whether
> > > we take #174 or not.
> >=20
> > See above. We already trust the leader to help produce a correct
> > result, but this design *also* trusts the intermediary to do so,
> > so now it's yet another trusted part of the system.
>=20
> I don=E2=80=99t really see this as a difference. The leader was =
already doing two functions: ingesting reports and aggregating. This =
proposal allows these functions to be split up without substantially =
altering the trust model.
>=20
> But this is precisely my point about the threat model: that the =
ingestor is not some kind of untrusted service, it is an integral part =
of the system, and has to be trusted, at least for the purpose of =
producing accurate results. So, now we need an analysis of what the =
intermediary can or cannot do.

Hmm, I think we may be talking past one another, or at least I may not =
be writing clearly, or I=E2=80=99m not understanding you. In any case, =
my claim is that the intermediary is just separating out the =
functionality that the Leader was already doing. And by your reasoning =
above, which I agree with, it would need to be trusted just like the =
Leader was trusted for ingestion. I don=E2=80=99t think the ingestion =
functionality can be considered an untrusted part of the system, for =
some of the reasons discussed above (batching considerations, report =
filtering, etc).

To try and restate my current position more clearly: I view use of an =
ingester as a deployment specific decision, and split upload should work =
such that it allows one to deploy an ingester if it=E2=80=99s useful to =
them.

>=20
> > >> - How do we deal with cases where the aggregators have =
inconsistent
> > >>   data, e.g., because an upload fails? Does this just fail? Do =
you
> > >>   retry? How do you know? Do we need some protocol to find out
> > >>   in advance so you don't get errors?
> > >
> > > This was discussed in detail in the previous thread, so I'll just
> > > leave a link to Tim's summary: [3]. The gist is that many of the =
new
> > > failure modes that arise are similar to failure modes that already
> > > exist, and we'll need to deal with these in much the same way. =
Thus
> > > dealing with split uploads doesn't appear to require too much
> > > additional plumbing. If we missed something in that discussion, =
please
> > > let us know!
> >=20
> > Yeah, I don't think this conclusion is correct. It's correct that
> > all of these failure modes exist to some extent with the current
> > model, but scale matters. In particular, with some reasonable
> > design choices (for instance, having the same retention policy
> > on each helper, to use Tim's example), you can reduce the number
> > of inconsistencies to a fairly small number when you can just report
> > it as error. However, it's quite plausible that in the environment
> > contemplated by #174 where each client uploads separately, you
> > see regular inconsistencies at a reasonable fraction (single digit
> > percentages?) of the number of reports, in which case you can't
> > just write them off as measurement error but actually need a system
> > to recover from them. This design doesn't provide such a recoevery
> > mechanism, AFAICT.
>=20
> Well, it _could_ bake in such recovery mechanisms, right? (I agree =
that more work is needed on the PR, but this doesn=E2=80=99t seem like a =
dealbreaker to me.)
>=20
> Those recovery mechanisms may well be quite complicated.  Given that =
the nominal purpose of this change is to reduce certain kinds of =
operational complexity, I'd like to understand what the resulting =
additional complexity of recovery is before making it.

Well, there=E2=80=99s at least two types of recovery here: recovery if =
the client fails to upload, and recovery between aggregators if they =
don=E2=80=99t agree on the shares. We already need to deal with the =
latter. What I was referring to was the former, which would be a new =
failure mode. As discussed in the other thread on this topic, the latter =
may be sufficient to address the former, or it may not be (I think your =
point about scale is important here). In any case, yeah, I think we can =
work through these mechanisms, especially the former.

Best,
Chris

>=20
> -Ekr
> =20
>=20
>=20
>=20
>=20
> Best,
> Chris
>=20
> >=20
> > -Ekr
> >=20
> >=20
> >=20
> >=20
> > --=20
> > Ppm mailing list
> > Ppm@ietf.org
> > https://www.ietf.org/mailman/listinfo/ppm
>=20


From nobody Thu Jan 27 16:13:10 2022
Return-Path: <rdd@cert.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id AF4C33A0E49 for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 16:13:07 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.8
X-Spam-Level: 
X-Spam-Status: No, score=-6.8 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, HTTP_ESCAPED_HOST=0.1, RCVD_IN_DNSWL_HI=-5, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=seicmu.onmicrosoft.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id PAaQbs_wROP8 for <ppm@ietfa.amsl.com>; Thu, 27 Jan 2022 16:13:05 -0800 (PST)
Received: from USG02-BN3-obe.outbound.protection.office365.us (mail-bn3usg02on0100.outbound.protection.office365.us [23.103.208.100]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 73F773A0E46 for <ppm@ietf.org>; Thu, 27 Jan 2022 16:13:05 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector5401; d=microsoft.com; cv=none; b=yyeQB7NOzlj4/oYZUb3dJvKbr7k+DPp5oAvI4ckR7HcOofQxhZo7LlsuHjvHOKY7Ma5N4VJL9pssV9j10gXnTA4tzB6ZhZjBZJeHbPKwDvMRpqw+I4HIntmjLpq0A4vWrY8jE6PnPvK3/7gpKs67BVXbJwCJ17YLEi0IHP4GJQC57jqEOf+0Oz1HCycTJ1BHGwlC+0gK0DDbGwKs4jmC1USlaCDaxhVMTM30kOvNROLlUYowrWeg02Vd24cqL34H/wXlOPAQTn1/5N3qYRzYzekelm78f71XhW0VrK1izirJQE887KxrvvTZTULkIS+0g8H1haTTIOMsBVdTwMy3zQ==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector5401; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=CBtWFPHdCZBhE9IBwynd4LnouEPqn8eYp8WWwfeia3s=; b=iqmX2HSBDPr5AN/4bnw/6p+TBVAheGfjJ+Z3Hpv6jeRv6GZeoih3+Bo5pu/RXQEJjD7T6jbztsVBjfy6N/qCohYM3b3aqR4/IPqD1TM8aFkzXV2QhAszK2hjGERKGg2RFnsEs+gRoCnchIFh9cgByjPUI5X5DihFhZqfaMZttf+huLfa0ouOabNY7ldHZdg7V6aRIxKchIqzIpgsK8Mm3yft1aW6uvTC/A+IU2CrnM+AEqUC6eEZ2wXYCfPsJGRbry3NTxFSWpVhTSB847yLmRXeyM1NEpQMCAb9W3y/jYJifU3i07TWTj74FYByUNFMI6SIqnvCp5vtMLrAWvbXow==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cert.org; dmarc=pass action=none header.from=cert.org; dkim=pass header.d=cert.org; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=seicmu.onmicrosoft.com; s=selector1-seicmu-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=CBtWFPHdCZBhE9IBwynd4LnouEPqn8eYp8WWwfeia3s=; b=KlkCet6+/D1aoxw50LgzKw/JOkbhSn/AmqiMvupiQpACVx0kADy9retIZ5APafDkqQ9UNs0Hu7ZCkVvD8cNkk6w7M920HGuYP9q5DQ5WDtAENaxu+5pI2ZWL7Z+PQnZPt+mYvP3+9NvK/PWW+NKVkvlBLrVTVXu6AObn1mlV2Ws=
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::11) by BN2P110MB1731.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::21) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4930.17; Fri, 28 Jan 2022 00:13:00 +0000
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e]) by BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e%3]) with mapi id 15.20.4930.018; Fri, 28 Jan 2022 00:13:00 +0000
From: Roman Danyliw <rdd@cert.org>
To: "ppm@ietf.org" <ppm@ietf.org>
Thread-Topic: Call for consensus on updated charter
Thread-Index: AdgCd8GCV7y9nJtBRKqSR9LxS59tQQFZVVmgAv6RryA=
Date: Fri, 28 Jan 2022 00:13:00 +0000
Message-ID: <BN2P110MB1107F5377F2B935E62AC5D7DDC229@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cert.org;
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: bb911fe6-687f-4fcd-2613-08d9e1f2f1e2
x-ms-traffictypediagnostic: BN2P110MB1731:EE_
x-microsoft-antispam-prvs: <BN2P110MB1731EA69618DC5598D66BC38DC229@BN2P110MB1731.NAMP110.PROD.OUTLOOK.COM>
x-ms-oob-tlc-oobclassifiers: OLM:10000;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM; PTR:; CAT:NONE;  SFS:(13230001)(366004)(83380400001)(498600001)(966005)(71200400001)(64756008)(66446008)(66476007)(66556008)(66946007)(76116006)(33656002)(8676002)(38070700005)(8936002)(9686003)(52536014)(86362001)(15650500001)(186003)(82960400001)(38100700002)(26005)(5660300002)(122000001)(6506007)(6916009)(7696005)(53546011)(2906002)(55016003)(20210929001); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: 8Em/KV1GDh7hF9v1iHjRg+DRCPYQohsNLrcaLmLwH/VF0hp5Wcl44+aXdvmmylHFoatdaj2OjMmMKR5YaheaqkJw7aQplH1oOttGA/IHJ7Coa/aLWcFCa82apRIhYghBBf5WOgaWRCR79k6VQETvJA==
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: cert.org
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM
X-MS-Exchange-CrossTenant-Network-Message-Id: bb911fe6-687f-4fcd-2613-08d9e1f2f1e2
X-MS-Exchange-CrossTenant-originalarrivaltime: 28 Jan 2022 00:13:00.3913 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 95a9dce2-04f2-4043-995d-1ec3861911c6
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN2P110MB1731
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/WZ_cI2yCbloER0dSJXQBamoRxeg>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 28 Jan 2022 00:13:08 -0000

Hi!

All of the input is appreciated.  Based on the stated interest demonstrated=
 here and previously at the BoF; and the lack of objections to the revised =
charter intended to address concerns raised at the PRIV BOF, we appear to h=
ave enough to get started with the WG chartering process.

As the next step, I'll get this charter information into the datatracker us=
ing the name PPM, get a meeting slot held for IETF 113 meeting, and proceed=
 to the initial IESG review.  If based on community or IESG review a WG is =
not chartered by IETF 113, we'll have a conversation on about whether to us=
e this time slot as another BOF.

A big thanks to Alissa Cooper and Adam Roach who's leadership as co-chairs =
of the PRIV BOF at IETF 112 allowed us to get this start.  I'd love to have=
 them stay on, but sadly, neither will be continuing on after this initial =
launch.

As we enter a new phase for this work, I'd like to point out what Ben and I=
 announced on SAAG [1] -- we are always looking for volunteers who would be=
 interested in chairing WGs in the future.  In particular, we'll be needing=
 chairs for the PPM _WG_.  If you are interested or have a recommendation, =
please reach out off-list.

Regards,
Roman

[1] https://mailarchive.ietf.org/arch/msg/saag/YL0LZJjxFZnHKNsaUPH3E-q1y4U/

> -----Original Message-----
> From: Roman Danyliw
> Sent: Wednesday, January 12, 2022 12:54 PM
> To: ppm@ietf.org
> Subject: RE: Call for consensus on updated charter
>=20
> Hi!
>=20
> Thanks to all that have responded to this email.
>=20
> A reminder that additional input is welcome and would be appreciated by
> Thursday, January 20.
>=20
> Regards,
> Roman
>=20
> > -----Original Message-----
> > From: Ppm <ppm-bounces@ietf.org> On Behalf Of Roman Danyliw
> > Sent: Wednesday, January 5, 2022 4:19 PM
> > To: ppm@ietf.org
> > Subject: [Ppm] Call for consensus on updated charter
> >
> > Hi!
> >
> > Happy New Year! Thanks for the robust discussion at the PRIV BOF at IET=
F
> 112.
> > The meeting notes [1] captured four major themes of feedback based on
> > the presented charter [2] (version 00-00).  Specifically:
> >
> > =3D=3D[ snip from meeting notes ]=3D=3D
> > (a) Generalize the text to make it clear that flexibility for
> > alternatives and additions to approach is possible (i.e., more than
> > Prio, heavy hitters)
> > (b) Refine the definition of aggregations so as to not restrict
> > alternatives
> > (c) Add a work item(s) to document the abuse cases and associated
> > mitigations from the perspective of all participants in the
> > architecture
> > (d) Revision of the WG name to convey a more narrow scope =3D=3D[ snip =
]=3D=3D
> >
> > In response to this feedback, the proposed charter text has been
> > revised at [3] (version 00-01).  Also, see [4] for a diff of this
> > revision against the original text presented at IETF 112.  For
> > convenience, this charter text is loaded into the datatracker using
> > the BOF name (PRIV). As there wasn't consensus on "PRIV" as the WG
> > name per feedback (d) or the now proposed alternative of "PPM", it
> > didn't make sense to ask the Secretariat to create a new entry in the
> > datatracker yet.  If consensus is found, a new datatracker entry will b=
e made
> prior to sending the charter to the IESG.
> >
> > This email is intended to start a thread to gauge interest in forming
> > a WG around this new, refined charter language [3].  Please respond to =
the
> list:
> >
> > (1) Do you support the charter text [3]? Or do you have objections or
> > blocking concerns (please describe what they might be)?
> >
> > If you do support the charter text:
> > (2) Are you willing to author or participate in the developed of the WG
> drafts?
> > (3) Are you willing to review the WG drafts?
> > (4) Are you interested in implementing the WG drafts?
> >
> > This call for consensus will end on Thursday, January 20.
> >
> > Regards,
> > Roman
> >
> > [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> > [2] charter-ietf-priv-00-00 -
> > https://datatracker.ietf.org/doc/charter-ietf-
> > priv/00-00/
> > [3] charter-ietf-priv-00-01 -
> > https://datatracker.ietf.org/doc/charter-ietf-
> > priv/00-01/
> > [4] diff of 00-00 vs. 00-01 --
> > https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.org%=
2
> > Fdoc
> > %2Fcharter-ietf-priv%2Fwithmilestones-00-
> > 00.txt&url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf-
> > priv%2Fwithmilestones-00-01.txt
> > [5] https://datatracker.ietf.org/wg/priv/about/
> >
> > --
> > Ppm mailing list
> > Ppm@ietf.org
> > https://www.ietf.org/mailman/listinfo/ppm


From nobody Sun Jan 30 00:16:01 2022
Return-Path: <do_not_reply@mnot.net>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 152103A1CFE for <ppm@ietfa.amsl.com>; Sun, 30 Jan 2022 00:15:58 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.698
X-Spam-Level: 
X-Spam-Status: No, score=-1.698 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_INVALID=0.1, DKIM_SIGNED=0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=no autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=fail (2048-bit key) reason="fail (message has been altered)" header.d=mnot.net header.b=W9MVErwC; dkim=fail (2048-bit key) reason="fail (message has been altered)" header.d=messagingengine.com header.b=ECamNiim
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id DtZQ4OxZvHlK for <ppm@ietfa.amsl.com>; Sun, 30 Jan 2022 00:15:53 -0800 (PST)
Received: from wout4-smtp.messagingengine.com (wout4-smtp.messagingengine.com [64.147.123.20]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 8396A3A1CFC for <ppm@ietf.org>; Sun, 30 Jan 2022 00:15:53 -0800 (PST)
Received: from compute5.internal (compute5.nyi.internal [10.202.2.45]) by mailout.west.internal (Postfix) with ESMTP id 6498E3200F9C for <ppm@ietf.org>; Sun, 30 Jan 2022 02:39:24 -0500 (EST)
Received: from mailfrontend2 ([10.202.2.163]) by compute5.internal (MEProxy); Sun, 30 Jan 2022 02:39:24 -0500
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=mnot.net; h=cc :content-type:date:from:from:in-reply-to:mime-version:reply-to :sender:subject:subject:to:to; s=fm1; bh=iGgtMBPvTxgwBhM57UVAfTH FL7bWn3q6brRiSx4C3CU=; b=W9MVErwCPr2aFLEOP3em8TXDsOscM6U8yrkhC5o KRFHpAbgD2KE4iOIvj8MEAxrEuH/f8k3/Ra7fpRGcM3U5XRsTMb1bnIBp9ZB+11J 2LDiAutnl8XKO7JIsUQo7PqEN5Z2blToIyX2R4W8XWnopAFyJTWOntGE4W0FmZBd rh6MEtyBwm13DZtwGAPkwrzJiDGJspM+Vbo/CtYpcG5jTw3lfJPEYutsqvyz4s+b bfWCDsZ8T4PgO0S40XZFVgRWpslrgGpKVSHSuLGrd8qdwnFRFboytZA3yZ9gJTyW yeEca2SBghQy8xUrXEn81FFAwm3e3keGhDxD02A+E/jk0PA==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-type:date:from:from :in-reply-to:mime-version:reply-to:sender:subject:subject:to:to :x-me-proxy:x-me-proxy:x-me-sender:x-me-sender:x-sasl-enc; s= fm1; bh=iGgtMBPvTxgwBhM57UVAfTHFL7bWn3q6brRiSx4C3CU=; b=ECamNiim JiJa3dNXn5n35PZCLo9v7lyjOZDuFM3BqUBjflhmRT0R736aFsWq60QQ14As2j5s YK9/ys2ITRyKuiH5Gg5UYJwTZS3jTuK/HifNChRMKy8wx5qxP5QRA59rWatrnk2A f6TeqHHl3HelqFQ5YOhJTkHcrqQQ/YEJbJKASJKPxigPKkySd1kLhlgwkqCiGeth 2qkKUBP4cN+7zX2c6nGGe9y0vQ9oJGQlzd1ZKMMnlTKi6SuPmekbWVNtItao+/8L 0lrpOSP3l5bYmtxsIOSRKk6EcxSgrtQDxfJaX6tHlWsPMKicajGk/Q3yd5x5vVUX 8l2LMcQitWNbUw==
X-ME-Sender: <xms:q0D2Yb_f2X8xba8AqQz-9DCUswmVjECipHpitlUMkZjNHIqnR2Bf0Q> <xme:q0D2YXv601wXaH4gDAxpVjaBzrNpgJXI_eO4FOpJW3jcw5gZ0Ya1hw8mq2olPPzg9 bURLYhZcARfPCY8tg>
X-ME-Received: <xmr:q0D2YZCAEITdlUkLWot8sxz8yZC_Ls-G4ntnMEa6JQNG_5CoDiyh_Ic8WsU3ef2sDeDjb0BYa-Sw8nv7Gpmk0kdVqh2qZdMACkcH3O1e02rPlYHtQWG0GjAhXy-h6ARZ1rY>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgedvvddrfeekgdeludcutefuodetggdotefrodftvf curfhrohhfihhlvgemucfhrghsthforghilhdpqfgfvfdpuffrtefokffrpgfnqfghnecu uegrihhlohhuthemuceftddtnecupfhoucgurghtvgcufhhivghlugculdegledmnecujf gurheptggghffvufesrgdttdertddtjeenucfhrhhomheptfgvphhoshhithhorhihucet tghtihhvihhthicuufhumhhmrghrhicuuehothcuoeguohgpnhhothgprhgvphhlhiesmh hnohhtrdhnvghtqeenucggtffrrghtthgvrhhnpeekfedvudetjedvfeekheeiveeugfef hfetteevgeffkefffeetffdvleehudeiteenucffohhmrghinhepghhithhhuhgsrdgtoh hmnecuvehluhhsthgvrhfuihiivgepgeenucfrrghrrghmpehmrghilhhfrhhomhepugho pghnohhtpghrvghplhihsehmnhhothdrnhgvth
X-ME-Proxy: <xmx:q0D2YXeJK7hudEuaUMvospcIUrd6UZwUWmUhzoGrbzPgvq-uCylXjA> <xmx:q0D2YQM2ArC6s20u42moX3ZuqLeCgTae7100TF1De1t0hrBS7AyiXQ> <xmx:q0D2YZlNItOVE3Wofg1b8AXEytIERK3wPn_cfQ_lHQgJQfUs31pUzQ> <xmx:q0D2YZbgDnjKW05GjVTFqnHz-tgwBGxAM5S3yB0DhhcnfpokBg8GwA>
Received: by mail.messagingengine.com (Postfix) with ESMTPA for <ppm@ietf.org>;  Sun, 30 Jan 2022 02:39:23 -0500 (EST)
Content-Type: multipart/alternative; boundary="===============8595216900648340428=="
MIME-Version: 1.0
From: Repository Activity Summary Bot <do_not_reply@mnot.net>
To: ppm@ietf.org
Message-Id: <20220130081553.8396A3A1CFC@ietfa.amsl.com>
Date: Sun, 30 Jan 2022 00:15:53 -0800 (PST)
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/R2qvTm5KhQ0UbcSmzaX__tlO0jU>
Subject: [Ppm] Weekly github digest (Privacy Preserving Measurement Activity Summary)
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 30 Jan 2022 08:15:58 -0000

--===============8595216900648340428==
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"; format="flowed"




Events=20

Issues
------
* abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC5)
  2 issues created:
  - Consider allowing clients to upload mulitple reports in a single messag=
e to aggregators (by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/188=20
  - Make sure error codes are enumerated in errors table (by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/187=20

  3 issues received 5 new comments:
  - #188 Consider allowing clients to upload mulitple reports in a single m=
essage to aggregators (2 by cjpatton, tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/188=20
  - #161 Rotation of shared secrets in PPM task parameters (2 by cjpatton, =
schoppmp)
    https://github.com/abetterinternet/ppm-specification/issues/161=20
  - #111 Do something about long-running requests (1 by tgeoghegan)
    https://github.com/abetterinternet/ppm-specification/issues/111=20

  1 issues closed:
  - Consider allowing clients to upload mulitple reports in a single messag=
e to aggregators https://github.com/abetterinternet/ppm-specification/issue=
s/188=20



Pull requests
-------------
* abetterinternet/ppm-specification (+0/-1/=F0=9F=92=AC0)
  1 pull requests merged:
  - Remove .targets.mk
    https://github.com/abetterinternet/ppm-specification/pull/186=20


Repositories tracked by this digest:
-----------------------------------
* https://github.com/abetterinternet/ppm-specification

--===============8595216900648340428==
Content-Type: text/html; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html lang=3D"en">
<head>
<meta charset=3D"utf-8">
<title>Weekly github digest (Privacy Preserving Measurement Activity Summar=
y)</title>
<style>
body { font-family: Gotham, "Helvetica Neue", Helvetica, Arial, sans-serif;=
 font-size: 14px; }
h2 { margin-top: 3em; color: #A52A2A; font-style: italic; font-weight: norm=
al; }
h3 { margin-bottom:0; margin-top: 2em; font-size: 1.2em; }
h1+h2 { margin-top: 1em; }
a { color: #bb6219; text-decoration: none; }
li { margin-bottom: .35em; }
.repos { margin-bottom: 0; margin-top:0; line-height: 1.2; }
.new { color: red; }
.label { display: inline;
	padding: .2em .6em .3em;
	font-size: 75%;
	font-weight: 700;
	line-height: 1;
	color: #fff;
	text-align: center;
	white-space: nowrap;
	vertical-align: baseline;
	border-radius: .25em;
}
details {
	margin-top: 8em;
	}
summary {
	margin-bottom: 1em;
	cursor: pointer;
}
</style>
</head>

<body>
<h1>Sunday January 30, 2022</h1>

<p>Events </p>

<h2>Issues</h2>

<h3>abetterinternet/ppm-specification (+2/-1/=F0=9F=92=AC5)</h3>
  <p class=3D"new">2 issues created:</p>
  <ul>
  <li>#188 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/188">Consider allowing clients to upload mulitple reports in a singl=
e message to aggregators</a> (by tgeoghegan) </li>
 =20
  <li>#187 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/187">Make sure error codes are enumerated in errors table</a> (by tg=
eoghegan) </li>
  </ul>

  <p>3 issues received 5 new comments:</p>
  <ul>
  <li>#188 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/188">Consider allowing clients to upload mulitple reports in a singl=
e message to aggregators</a> (2 by cjpatton, tgeoghegan) </li>
 =20
  <li>#161 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/161">Rotation of shared secrets in PPM task parameters</a> (2 by cjp=
atton, schoppmp) </li>
 =20
  <li>#111 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/111">Do something about long-running requests</a> (1 by tgeoghegan) =
</li>
  </ul>

  <p>1 issues closed:</p>
  <ul>
  <li>#188 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
issues/188">Consider allowing clients to upload mulitple reports in a singl=
e message to aggregators</a> </li>
  </ul>



<h2>Pull requests</h2>
<h3>abetterinternet/ppm-specification (+0/-1/=F0=9F=92=AC0)</h3>


  <p>1 pull requests merged:</p>
  <ul>
  <li>#186 <a href=3D"https://github.com/abetterinternet/ppm-specification/=
pull/186">Remove .targets.mk</a> </li>
  </ul>


  <details>
    <summary>Repositories tracked by this digest:</summary>
<ul class=3D"repos">
  <li><a href=3D"https://github.com/abetterinternet/ppm-specification">http=
s://github.com/abetterinternet/ppm-specification</a></li>
</ul>
</details>
</body>
</html>

--===============8595216900648340428==--


From nobody Mon Jan 31 10:55:42 2022
Return-Path: <rdd@cert.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 5F36E3A132C for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 10:55:40 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.9
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=seicmu.onmicrosoft.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id sHZ6nv-CSCgH for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 10:55:36 -0800 (PST)
Received: from USG02-BN3-obe.outbound.protection.office365.us (mail-bn3usg02on0094.outbound.protection.office365.us [23.103.208.94]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id C41B43A08A5 for <ppm@ietf.org>; Mon, 31 Jan 2022 10:55:35 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector5401; d=microsoft.com; cv=none; b=TL6IEbKpNv5xjBv8z3WpwEYdRW1M3ZNmw4hB3XN8Nz3uGlbQ24aiJpiZbUxQpaKjGrpkMpWaGJZJu3K6j52DUbPsNC2qA8vaOD387Lg9L8Skx1lNCMn9CKFKCB6nCvF7CaSPWP1TCF+E8nD4inDnQxob2IEVw2APVVwksCVPWgol1GkwTmxYw8qbS0km4CC80VTZ4CTtNG6MfaPlqvrjvA5i1y3PmLQ6gldyKe2TSaOluZANgb1+b6wZt92T4dZhN11r26bfItZtFxDLSb/4Ue+l9sanZt4/PqLQZAs/jk+Du3r3jaRs4uXoNxRhB+FH10Q1NdD4kgDF2PH8vDFgeQ==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector5401; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=MC3akZ1Zf2GeM/DK9zDuk0iN0eGru9YvDXx+x0FRw8w=; b=Bz4RADcCZ84BlflI99nW7pQIrI86KGfL4MNDHtXki8X3HEOg3GDXb05vIKAvoaiPy0jAepAh72/FF0bv+Y9zQsvBou9GP16iwrp5EOu4dTPZOMcYhT9EQJ9oOmOSRw/3pUwiwcIPKqlMbijsCwafZCkfF8t15KIr4CfcJ19XQCt5mwEM4kA4DFVnqzOgj6P3Elg5J4RvEgdYu2u6Xx9DXGfXXQNmSqTyQb3Y6i6dYB/O/1ML3b2rSRFBWn3AU3846EsbdwzpJ4iftVCgaqGBkWV96ADSgy3O8XCfEpQOGRabr6xY0ClqLOp9aOULPVK5YDcrc9zhZQJDg9YsMPbAeQ==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cert.org; dmarc=pass action=none header.from=cert.org; dkim=pass header.d=cert.org; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=seicmu.onmicrosoft.com; s=selector1-seicmu-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=MC3akZ1Zf2GeM/DK9zDuk0iN0eGru9YvDXx+x0FRw8w=; b=g0SVBgY0aoB+ecYUl9Hi3l81xYIAgvIMFMQUaNQJuV0+OHyV2/gEGY5X9y/MFuhX/s5iKZPnMJr30VfqHH2cPn18Aws1HQvNa76L8SIm70VoeSBA7kaF8CQRHCjq8gqtRbOhpIS8VfqLeHw5H/p0C2CPeDmOwcwpU5xtpnuvfo4=
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::11) by BN2P110MB1351.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:17c::16) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4930.21; Mon, 31 Jan 2022 18:55:28 +0000
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e]) by BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e%3]) with mapi id 15.20.4930.020; Mon, 31 Jan 2022 18:55:28 +0000
From: Roman Danyliw <rdd@cert.org>
To: "ppm@ietf.org" <ppm@ietf.org>
Thread-Topic: Call for consensus on updated charter
Thread-Index: AdgCd8GCV7y9nJtBRKqSR9LxS59tQQFZVVmgAv6RryAAvq4OAA==
Date: Mon, 31 Jan 2022 18:55:28 +0000
Message-ID: <BN2P110MB1107578AB2C29500F1A4D169DC259@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
References: <BN2P110MB11074BC9720E7D644097D53BDC4B9@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <BN2P110MB110729FC433161F84C73627DDC529@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM> <BN2P110MB1107F5377F2B935E62AC5D7DDC229@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB1107F5377F2B935E62AC5D7DDC229@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cert.org;
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 2b4babd5-4c5a-4316-168e-08d9e4eb3f9d
x-ms-traffictypediagnostic: BN2P110MB1351:EE_
x-microsoft-antispam-prvs: <BN2P110MB13513EB82C29968A99FFFC12DC259@BN2P110MB1351.NAMP110.PROD.OUTLOOK.COM>
x-ms-oob-tlc-oobclassifiers: OLM:7691;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM; PTR:; CAT:NONE;  SFS:(13230001)(366004)(66446008)(2906002)(66946007)(86362001)(66476007)(66556008)(122000001)(498600001)(76116006)(8936002)(64756008)(8676002)(71200400001)(52536014)(83380400001)(55016003)(5660300002)(38070700005)(33656002)(38100700002)(9686003)(53546011)(966005)(7696005)(186003)(26005)(6506007)(82960400001)(15650500001)(6916009)(20210929001); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: tm2rQCfW1UN2N6o+iDzZiAXcpHxt6uI+ieFc3eb66nHKy9L9AE7OvWwI6wTbRVps3QpXb6NRNCy93FWdvok55J8Hc/UcOB2JdOV/ipVx3jY96E142Utad0quPLn8HUA5OUEzek1V13Zt8yWeC1urow==
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: cert.org
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM
X-MS-Exchange-CrossTenant-Network-Message-Id: 2b4babd5-4c5a-4316-168e-08d9e4eb3f9d
X-MS-Exchange-CrossTenant-originalarrivaltime: 31 Jan 2022 18:55:28.4717 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 95a9dce2-04f2-4043-995d-1ec3861911c6
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN2P110MB1351
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/JygLgzcv1FYnPzO3QXFbozoH4Q8>
Subject: Re: [Ppm] Call for consensus on updated charter
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 31 Jan 2022 18:55:40 -0000

Hi!

The charter we discussed is now staged under the new WG name, PPM, at https=
://datatracker.ietf.org/wg/ppm/about/.

One last bit of charter business -- we need to define milestone(s) associat=
ed with the charter.  My initial proposal would be something of the form:

=3D=3D[ snip ]=3D=3D
Dec 2023   Submit PPM protocol to IESG for publication
=3D=3D[ snip ]=3D=3D

Feedback welcome.  Please keep in mind this is a rough estimate (both on th=
e timing and number of deliverables) to get the WG off the ground.  Should =
a WG be approved, this can be updated and refined.  For example, the charte=
r scope allows for the possibility of multiple protocols and different PPM =
algorithms.

Regards,
Roman



> -----Original Message-----
> From: Ppm <ppm-bounces@ietf.org> On Behalf Of Roman Danyliw
> Sent: Thursday, January 27, 2022 7:13 PM
> To: ppm@ietf.org
> Subject: Re: [Ppm] Call for consensus on updated charter
>=20
> Hi!
>=20
> All of the input is appreciated.  Based on the stated interest demonstrat=
ed here
> and previously at the BoF; and the lack of objections to the revised char=
ter
> intended to address concerns raised at the PRIV BOF, we appear to have
> enough to get started with the WG chartering process.
>=20
> As the next step, I'll get this charter information into the datatracker =
using the
> name PPM, get a meeting slot held for IETF 113 meeting, and proceed to th=
e
> initial IESG review.  If based on community or IESG review a WG is not
> chartered by IETF 113, we'll have a conversation on about whether to use =
this
> time slot as another BOF.
>=20
> A big thanks to Alissa Cooper and Adam Roach who's leadership as co-chair=
s of
> the PRIV BOF at IETF 112 allowed us to get this start.  I'd love to have =
them stay
> on, but sadly, neither will be continuing on after this initial launch.
>=20
> As we enter a new phase for this work, I'd like to point out what Ben and=
 I
> announced on SAAG [1] -- we are always looking for volunteers who would b=
e
> interested in chairing WGs in the future.  In particular, we'll be needin=
g chairs
> for the PPM _WG_.  If you are interested or have a recommendation, please
> reach out off-list.
>=20
> Regards,
> Roman
>=20
> [1] https://mailarchive.ietf.org/arch/msg/saag/YL0LZJjxFZnHKNsaUPH3E-
> q1y4U/
>=20
> > -----Original Message-----
> > From: Roman Danyliw
> > Sent: Wednesday, January 12, 2022 12:54 PM
> > To: ppm@ietf.org
> > Subject: RE: Call for consensus on updated charter
> >
> > Hi!
> >
> > Thanks to all that have responded to this email.
> >
> > A reminder that additional input is welcome and would be appreciated
> > by Thursday, January 20.
> >
> > Regards,
> > Roman
> >
> > > -----Original Message-----
> > > From: Ppm <ppm-bounces@ietf.org> On Behalf Of Roman Danyliw
> > > Sent: Wednesday, January 5, 2022 4:19 PM
> > > To: ppm@ietf.org
> > > Subject: [Ppm] Call for consensus on updated charter
> > >
> > > Hi!
> > >
> > > Happy New Year! Thanks for the robust discussion at the PRIV BOF at
> > > IETF
> > 112.
> > > The meeting notes [1] captured four major themes of feedback based
> > > on the presented charter [2] (version 00-00).  Specifically:
> > >
> > > =3D=3D[ snip from meeting notes ]=3D=3D
> > > (a) Generalize the text to make it clear that flexibility for
> > > alternatives and additions to approach is possible (i.e., more than
> > > Prio, heavy hitters)
> > > (b) Refine the definition of aggregations so as to not restrict
> > > alternatives
> > > (c) Add a work item(s) to document the abuse cases and associated
> > > mitigations from the perspective of all participants in the
> > > architecture
> > > (d) Revision of the WG name to convey a more narrow scope =3D=3D[ sni=
p
> > > ]=3D=3D
> > >
> > > In response to this feedback, the proposed charter text has been
> > > revised at [3] (version 00-01).  Also, see [4] for a diff of this
> > > revision against the original text presented at IETF 112.  For
> > > convenience, this charter text is loaded into the datatracker using
> > > the BOF name (PRIV). As there wasn't consensus on "PRIV" as the WG
> > > name per feedback (d) or the now proposed alternative of "PPM", it
> > > didn't make sense to ask the Secretariat to create a new entry in
> > > the datatracker yet.  If consensus is found, a new datatracker entry
> > > will be made
> > prior to sending the charter to the IESG.
> > >
> > > This email is intended to start a thread to gauge interest in
> > > forming a WG around this new, refined charter language [3].  Please
> > > respond to the
> > list:
> > >
> > > (1) Do you support the charter text [3]? Or do you have objections
> > > or blocking concerns (please describe what they might be)?
> > >
> > > If you do support the charter text:
> > > (2) Are you willing to author or participate in the developed of the
> > > WG
> > drafts?
> > > (3) Are you willing to review the WG drafts?
> > > (4) Are you interested in implementing the WG drafts?
> > >
> > > This call for consensus will end on Thursday, January 20.
> > >
> > > Regards,
> > > Roman
> > >
> > > [1] https://datatracker.ietf.org/doc/minutes-112-priv/
> > > [2] charter-ietf-priv-00-00 -
> > > https://datatracker.ietf.org/doc/charter-ietf-
> > > priv/00-00/
> > > [3] charter-ietf-priv-00-01 -
> > > https://datatracker.ietf.org/doc/charter-ietf-
> > > priv/00-01/
> > > [4] diff of 00-00 vs. 00-01 --
> > > https://www.ietf.org/rfcdiff?url1=3Dhttps%3A%2F%2Fdatatracker.ietf.or=
g
> > > %2
> > > Fdoc
> > > %2Fcharter-ietf-priv%2Fwithmilestones-00-
> > > 00.txt&url2=3Dhttps%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fcharter-ietf=
-
> > > priv%2Fwithmilestones-00-01.txt
> > > [5] https://datatracker.ietf.org/wg/priv/about/
> > >
> > > --
> > > Ppm mailing list
> > > Ppm@ietf.org
> > > https://www.ietf.org/mailman/listinfo/ppm
>=20
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm


From nobody Mon Jan 31 11:48:33 2022
Return-Path: <session-request@ietf.org>
X-Original-To: ppm@ietf.org
Delivered-To: ppm@ietfa.amsl.com
Received: from ietfa.amsl.com (localhost [IPv6:::1]) by ietfa.amsl.com (Postfix) with ESMTP id F16453A1538; Mon, 31 Jan 2022 11:48:31 -0800 (PST)
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
From: IETF Meeting Session Request Tool <session-request@ietf.org>
To: <session-request@ietf.org>
Cc: ppm-chairs@ietf.org, ppm@ietf.org, rdd@cert.org
X-Test-IDTracker: no
X-IETF-IDTracker: 7.44.0
Auto-Submitted: auto-generated
Precedence: bulk
Message-ID: <164365851189.11948.14352399559163112334@ietfa.amsl.com>
Date: Mon, 31 Jan 2022 11:48:31 -0800
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/0wOcz6Z42R-cLt3vcJymRC4veXM>
Subject: [Ppm] ppm - New Meeting Session Request for IETF 113
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 31 Jan 2022 19:48:32 -0000

A new meeting session request has just been submitted by Roman Danyliw, a SEC Area Director.


---------------------------------------------------------
Working Group Name: Privacy Preserving Measurement
Area Name: Security Area
Session Requester: Roman Danyliw


Number of Sessions: 1
Length of Session(s): 2 Hours
Number of Attendees: 75
Conflicts to Avoid: 
 Technology overlap: saag
 Key participant conflict: tls cfrg

       


People who must be present:
  Roman Danyliw

Resources Requested:

Special Requests:
  
---------------------------------------------------------



From nobody Mon Jan 31 17:10:24 2022
Return-Path: <rdd@cert.org>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 77A0C3A0981 for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 17:10:22 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.9
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=seicmu.onmicrosoft.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id rPar4VAnyYOa for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 17:10:18 -0800 (PST)
Received: from USG02-CY1-obe.outbound.protection.office365.us (mail-cy1usg02on0123.outbound.protection.office365.us [23.103.209.123]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id F0DEC3A0983 for <ppm@ietf.org>; Mon, 31 Jan 2022 17:10:17 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; s=arcselector5401; d=microsoft.com; cv=none; b=H1FiV4M1dBaupNPzR3ZSNrghuwMwDinH4mBxoSwsmDAjjAQSt4xs5gNFQF/aAuvi49i/mOnGKE0tlk8d8SCAEbbE2MHZmTm4UYlnEO/zofhP26Hb4X2QU7e9n8Q12SYwqcQxlyMiAvVXGigZuFb2q8izxR31GBK9LYV5A08/2Qh+Bg1VB0C3E09161yynBkIoi5Pr0D1YN0e7x6UHQMMdezhiKFiSHn3aIbhxkruKL55uD5/VctB+a3BlHgMuNC/XAtPdDjWv1dD85VuoDWydQsqdCl1yhpyNQsdXRU0urWkorS8KwqWEvpSgToI0mNA+bo/F72jHnekDW9gJKFzSw==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector5401; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=pBFRXQVORSqMrvqWD9Rp9E30/mPyiRK43DkUKICcTEQ=; b=A7jp741fmauLcpeO6q/KzWwia+PhXyg/bJxaBe3hd8tpUbJtVUbM97OBTSeJWqth85DLPMJ3Nd1F/xN9r/EV9fAqLpSaPZzbLSm70kSjBQTKZ8cot7BBiEivRFkhF/F5FepcgJCTtBGKweeQRihd30IvE6qWdaYjmpsaqwzF0Crpi0aEwNP75WbUYFVD5Pf1A3/xIbyo4NHxlDMjqEOOG2QwsYZqNS3E6NWtvA0Dp2XIbS4H7fvFOzL6dNwuW6rzWBA81MessnThkkpAuR0itqamMXjoI4+BXpnxjF3sMSNap9h6NUedWH1uNgWyEsm8OdmhumhjpiQ0bahsbS0GPg==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=cert.org; dmarc=pass action=none header.from=cert.org; dkim=pass header.d=cert.org; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=seicmu.onmicrosoft.com; s=selector1-seicmu-onmicrosoft-com; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=pBFRXQVORSqMrvqWD9Rp9E30/mPyiRK43DkUKICcTEQ=; b=CyNFFvMYFFLyVCxDelKZ71osagKVkneHOSynxa58GGEfrIil15PW1BOJooAShqjyek4127FyQEyHIt54yyi2WDjGFszC3pkRW75Z1tS0SJwn77D4O4rjcg2EYOPaxd7vG0Odqe8xnKzwroppjCqxAAmL06DCTmp6kEDG4EwN794=
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:168::11) by BN2P110MB1238.NAMP110.PROD.OUTLOOK.COM (2001:489a:200:17f::5) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4930.17; Tue, 1 Feb 2022 01:10:13 +0000
Received: from BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e]) by BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM ([fe80::45f0:b470:9c74:ef6e%3]) with mapi id 15.20.4930.020; Tue, 1 Feb 2022 01:10:13 +0000
From: Roman Danyliw <rdd@cert.org>
To: "ppm@ietf.org" <ppm@ietf.org>
Thread-Topic: New chair for PPM
Thread-Index: AdgXCA1bTZQKt6T8Q+S9mjsfMZsaTA==
Date: Tue, 1 Feb 2022 01:10:13 +0000
Message-ID: <BN2P110MB11071770EEDC7F080713AF7EDC269@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=cert.org;
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 34d10106-a362-4a8b-1081-08d9e51f99b4
x-ms-traffictypediagnostic: BN2P110MB1238:EE_
x-microsoft-antispam-prvs: <BN2P110MB1238F2FAE5A1FCE31B301429DC269@BN2P110MB1238.NAMP110.PROD.OUTLOOK.COM>
x-ms-oob-tlc-oobclassifiers: OLM:6790;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM; PTR:; CAT:NONE;  SFS:(13230001)(366004)(122000001)(82960400001)(86362001)(38070700005)(38100700002)(3480700007)(52536014)(66476007)(66446008)(66946007)(8676002)(8936002)(33656002)(76116006)(66556008)(55016003)(4744005)(5660300002)(64756008)(9686003)(2906002)(7696005)(6506007)(26005)(186003)(6916009)(71200400001)(966005)(498600001)(20210929001); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: LA8rjT7/U2jRyHG4xE8YRgx9xciUvsNvv78K0zpRhl9833jNB7IIJprmpse02AytBweFGcSSykgpQ3k4oSOx78ytaD+a8smntFeY5SXEJ2MB7DOtzkJQUblXmEcOfJi11wTzlCcEjp+uFsF3VRmi0Q==
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: cert.org
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM
X-MS-Exchange-CrossTenant-Network-Message-Id: 34d10106-a362-4a8b-1081-08d9e51f99b4
X-MS-Exchange-CrossTenant-originalarrivaltime: 01 Feb 2022 01:10:13.4315 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 95a9dce2-04f2-4043-995d-1ec3861911c6
X-MS-Exchange-Transport-CrossTenantHeadersStamped: BN2P110MB1238
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/nii3cNgasatiZUTME8v1za8eYTs>
Subject: [Ppm] New chair for PPM
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 01 Feb 2022 01:10:23 -0000

Hi!

As I noted earlier this week [1], PPM is entering a new phase as it is begi=
nning it's path to being a WG.  To help with that process and our future wo=
rk, it's my pleasure to welcome Ben Schwartz as the first of two chair to b=
e named for PPM.

Ben: thank you for your willingness to help!

Regards,
Roman

[1] https://mailarchive.ietf.org/arch/msg/ppm/WZ_cI2yCbloER0dSJXQBamoRxeg/


From nobody Mon Jan 31 19:42:54 2022
Return-Path: <cpatton@cloudflare.com>
X-Original-To: ppm@ietfa.amsl.com
Delivered-To: ppm@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 8DFBA3A102B for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 19:42:49 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.099
X-Spam-Level: 
X-Spam-Status: No, score=-7.099 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_HI=-5, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=cloudflare.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id kkIMSu9gxM_L for <ppm@ietfa.amsl.com>; Mon, 31 Jan 2022 19:42:44 -0800 (PST)
Received: from mail-pf1-x42e.google.com (mail-pf1-x42e.google.com [IPv6:2607:f8b0:4864:20::42e]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 7279C3A1029 for <ppm@ietf.org>; Mon, 31 Jan 2022 19:42:44 -0800 (PST)
Received: by mail-pf1-x42e.google.com with SMTP id i30so14661798pfk.8 for <ppm@ietf.org>; Mon, 31 Jan 2022 19:42:44 -0800 (PST)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=cloudflare.com; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=q4UuffYuPCWuUukDLHzm8JOv05z+osHQO6MgOQTu4ns=; b=iUHeIMU8dEZaDO9PUQilDx2iEMa0ivzgwl9NQLymxmVuHdIf0Wj5EAKQyi9xZV4UHs RrU11H+Fp5Gmklog6k1IYiqXcaEm32Wc/IdKdrZDPEjyk/G+ULRkTFpz0vXlG+SDOU5H SCzXj5gFAlCGoUr/b+KhQHt9uU6vUkrS338f4=
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=q4UuffYuPCWuUukDLHzm8JOv05z+osHQO6MgOQTu4ns=; b=g7GgDN79PMt+p3NSSd864tF36ndWmVARd6O97a+VFeJ17xlbEWDN1LEmH+HQetMved PgY8uOGviXTtTuqEQ1vMinF2iRvHYDrvLsKwLypW2mNFKWuRE925Le56dXvXKZw3JCaj kCoh1TJJwK6I62l+wYHnXgiLzVhv8+HgKmlgCdZoWs7YvMZg8d+6uhvltJkOcYe1cYtX wx48UusVl0liE3PlGuiN7He3ubxLpvMdZrwKZkm21CVKo6vvK6WAGUh+JRgtJ4yt/KqD eVHyYeQRc69VLOgfvDZUxIIitwCexACr819h7PHhNmdoZPnyBA2DHakobHX7n0SAaQpF 1dyg==
X-Gm-Message-State: AOAM530m0i8kafQ04NsS3j5GyeQO9Flxq1v54MWJizygXfN8OKvoSWlE W+NgKuOqJ1VIr8jxzC0LshuVAtvHSR67SxWoAItZ2hr1efQ=
X-Google-Smtp-Source: ABdhPJxjgiysbl6Z5JeW220wYEFdkOo9x86/1bvAlnEgoti1lKNU3pWCY6xOvjf+VOmGHZ2CImTAIDpoEoTeqMOGwOI=
X-Received: by 2002:a63:285:: with SMTP id 127mr19047685pgc.67.1643686962994;  Mon, 31 Jan 2022 19:42:42 -0800 (PST)
MIME-Version: 1.0
References: <BN2P110MB11071770EEDC7F080713AF7EDC269@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
In-Reply-To: <BN2P110MB11071770EEDC7F080713AF7EDC269@BN2P110MB1107.NAMP110.PROD.OUTLOOK.COM>
From: Christopher Patton <cpatton@cloudflare.com>
Date: Mon, 31 Jan 2022 19:42:31 -0800
Message-ID: <CAG2Zi23irWxfOkW0hNcsTbFVaZuTDh=vnAwU-1Z7AvX2oC2tKQ@mail.gmail.com>
To: Roman Danyliw <rdd@cert.org>
Cc: "ppm@ietf.org" <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="0000000000008d88c005d6ecb10f"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/BbNRtlxTyodiaY6XC_zzxs7McFI>
Subject: Re: [Ppm] New chair for PPM
X-BeenThere: ppm@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Privacy Preserving Measurement technologies <ppm.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/ppm>, <mailto:ppm-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/ppm/>
List-Post: <mailto:ppm@ietf.org>
List-Help: <mailto:ppm-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/ppm>, <mailto:ppm-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 01 Feb 2022 03:42:50 -0000

--0000000000008d88c005d6ecb10f
Content-Type: text/plain; charset="UTF-8"

Awesome, thanks Ben!

Best,
Chris P.

On Mon, Jan 31, 2022 at 5:10 PM Roman Danyliw <rdd@cert.org> wrote:

> Hi!
>
> As I noted earlier this week [1], PPM is entering a new phase as it is
> beginning it's path to being a WG.  To help with that process and our
> future work, it's my pleasure to welcome Ben Schwartz as the first of two
> chair to be named for PPM.
>
> Ben: thank you for your willingness to help!
>
> Regards,
> Roman
>
> [1] https://mailarchive.ietf.org/arch/msg/ppm/WZ_cI2yCbloER0dSJXQBamoRxeg/
>
> --
> Ppm mailing list
> Ppm@ietf.org
> https://www.ietf.org/mailman/listinfo/ppm
>

--0000000000008d88c005d6ecb10f
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Awesome, thanks Ben!</div><div><br></div><div>Best,</=
div><div>Chris P.<br></div></div><br><div class=3D"gmail_quote"><div dir=3D=
"ltr" class=3D"gmail_attr">On Mon, Jan 31, 2022 at 5:10 PM Roman Danyliw &l=
t;<a href=3D"mailto:rdd@cert.org">rdd@cert.org</a>&gt; wrote:<br></div><blo=
ckquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex">Hi!<br>
<br>
As I noted earlier this week [1], PPM is entering a new phase as it is begi=
nning it&#39;s path to being a WG.=C2=A0 To help with that process and our =
future work, it&#39;s my pleasure to welcome Ben Schwartz as the first of t=
wo chair to be named for PPM.<br>
<br>
Ben: thank you for your willingness to help!<br>
<br>
Regards,<br>
Roman<br>
<br>
[1] <a href=3D"https://mailarchive.ietf.org/arch/msg/ppm/WZ_cI2yCbloER0dSJX=
QBamoRxeg/" rel=3D"noreferrer" target=3D"_blank">https://mailarchive.ietf.o=
rg/arch/msg/ppm/WZ_cI2yCbloER0dSJXQBamoRxeg/</a><br>
<br>
-- <br>
Ppm mailing list<br>
<a href=3D"mailto:Ppm@ietf.org" target=3D"_blank">Ppm@ietf.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/ppm" rel=3D"noreferrer" ta=
rget=3D"_blank">https://www.ietf.org/mailman/listinfo/ppm</a><br>
</blockquote></div>

--0000000000008d88c005d6ecb10f--

