[Ppm] Agenda for IETF 114

Christopher Patton <cpatton@cloudflare.com> Tue, 19 July 2022 17:33 UTC

MIME-Version: 1.0
From: Christopher Patton <cpatton@cloudflare.com>
Date: Tue, 19 Jul 2022 10:33:12 -0700
Message-ID: <CAG2Zi20uuGQjR5FTg1E2NEM=BKo7pSidMKnU5rHU3AhTpqRYUw@mail.gmail.com>
To: ppm <ppm@ietf.org>
Content-Type: multipart/alternative; boundary="00000000000092bc9e05e42be195"
Archived-At: <https://mailarchive.ietf.org/arch/msg/ppm/J_0i3c8siw9OGihkGpmFoHXT-os>
Subject: [Ppm] Agenda for IETF 114
Precedence: list

Hi Aggregators, we have yet to settle on an agenda for IETF. The DAP (
https://datatracker.ietf.org/doc/html/draft-ietf-ppm-dap-01) editors have a
set of things they'd like to discuss, and there is a new STAR draft to
consider for adoption (
https://datatracker.ietf.org/doc/html/draft-dss-star-01). Does anyone else
have anything that would be useful to discuss as a group?

Things to discuss for DAP, in order of priority:

   1. Perhaps the most pressing question is whether the current mechanism
   for sorting reports into batches is sufficient to support our current use
   cases. We believe it's not and that changes to the collect sub-protocol are
   needed. To guide design, it would be useful to discuss the use cases people
   have and how to accommodate them. (See
   https://mailarchive.ietf.org/arch/msg/ppm/NKPIIm5HvZ1p8EkS3tmwj6DBXDs/
   for details.)
   2. Related to this question are our privacy considerations for the
   collect sub-protocol (
   https://github.com/ietf-wg-ppm/draft-ietf-ppm-dap/issues/195). In short,
   if batches are allowed to overlap, then aggregate results may end up
   leaking more information about individual reports than what the system
   intends. We'd like to discuss how mechanisms like differential privacy can
   help address this problem.
   3. Apart from (2.), we know that we're going to need to make
   accommodations for differential privacy in the DAP protocol. There are some
   interesting open questions to address here, starting with the mechanism
   itself: Can/should a mechanism for DP be made generic, or is this
   inherently dependent upon the underlying VDAF? (See
   https://github.com/cfrg/draft-irtf-cfrg-vdaf/issues/94 for discussion.)
   4. Lastly, we'd like to highlight some of the changes between DAP-00 and
   DAP-01 and point to some of the open issues we'd like to address by DAP-02.

Finally, for those who are not in the #ppm channel of the IETF slack: There
are two implementations of DAP-01 of which we are aware:

   - Daphne (https://github.com/cloudflare/daphne)
   - Janus (https://github.com/divviup/janus)

Both are written Rust and use the VDAFs implemented in
https://github.com/divviup/libprio-rs. We're eager to help get other
implementers up to speed, particularly in different languages!

Happy aggregating,
Chris P.

[Ppm] Agenda for IETF 114 Christopher Patton