By "double-blind" I mean that not only does the issuer not know who the client is, but the server doesn't know exactly which issuers the client redeemed trust tokens from.
Summary: Two ways for an untrusted client to prove to an untrusted server that the client is trusted by X out of Y trust token issuers, without revealing exactly which issuers:
- One way uses hash functions as its only crypto primitive, but requires revealing some of the issuers to the server, which is a leak of fingerprintable info. It's also probabilistic, and the chances of a malicious client's lie being caught depends on the degree of the lie as well as how many issuers are revealed to the server, which is an undesirable tradeoff.
- The other way, if it works, reveals none of the issuers to the server, and the client cannot lie at all. But this assumes I understand elliptic curve cryptography, which I learned entirely