What parties will participate in helper party networks?

[eriktaubeneck]

This issue was initially raised by @alexWhitworth on the meetings issue for Agenda Request - IPA Status Update, which essentially came down to:

Yes, I understood everything in your comment/proposal except whom the 3 parties are, particularly as it relates to a real-world use case.

I responded that the technical proposal is intentionally abstract about this, and that this is more a policy question. So this issue is intended to start a conversation around that policy question:

What parties will participate in helper party networks?

Ultimately, we both need to:

1. Reach consensus in the PATCG around what a helper party (or, in our Threat Model draft, an Aggregator) is, and what types of parties could be considered.
2. Find parties to form networks who the implementers (web platform vendors) will trust to uphold the assumption of non-collusion required from \1.

A notable effort in this type of area is DivviUp from the Internet Security Research Group (ISRG).

In our meeting on 8/11, @chris-wood mentioned that Cloudflare may be interested in such a role. (As @martinthomson pointed out in the other thread, it would be great if @chris-wood would help qualify this more precisely.)

Also, @alexcone pointed out that it's "hard [for him] to believe that public cloud provider would turn down digital advertising compute."

It would be great to continue this discussion here, and hear from folks both about both issues above: what properties these parties should have, and what configurations of helper party networks could be trusted to be non-colluding.

[chris-wood]

Thanks @eriktaubeneck. Cloudflare is definitely interested in experimenting here (as we are doing on the IETF-side of things for DAP).

[ohpauleez]

Kevel is ready to step up to be a helper node, utilizing the event processing infrastructure we already have in place and adapting it to handle the MPC protocol.
We've been eagerly waiting for the Working Group and POC work. We're committed to seeing IPA move forward and will support it however we can.

(Apologies for missing the meeting the other day where the MVP was discussed -- I'm still catching up).

[bmayd]

@eriktaubeneck can you provide a link to the thread referred to?

As @martinthomson pointed out in the other thread...

[eriktaubeneck]

@bmayd it's patcg/meetings#70, linked at the beginning of my issue.

[tgeoghegan]

To echo @eriktaubeneck's comment in the issue description, ISRG is building Divvi Up in support of lots of private measurement and metrics use cases. Divvi Up implements the Distributed Aggregation Protocol, which coordinates the execution of Verifiable Distributed Aggregation Functions. Besides the Prio3 family, we plan to support Poplar1, which among other things, can be used to solve some of the problems PATCG is addressing. We would be happy to have Divvi Up's Poplar1 support be a part of these deployments.

As @chris-wood mentioned, the standardization of this work is ongoing in the IETF's PPM working group, and we'd welcome input from PATCG experts to help us make sure that Poplar1 and DAP meet the needs of these emerging browser features.