Privacy interoperability research proposal
Status: proposal only. No credential, proof, ledger, or revocation protocol is changed by this document. Preserve U-net's locally owned relationship lifecycle while testing standards adapters before inventing another provider ecosystem.
Candidate approaches
| Approach | Useful property | Questions the prototype must answer |
|---|---|---|
| SD-JWT | Selective disclosure using JWT-based mechanisms | Does repeat use expose stable signed material, holder keys, or status references? Do not treat selective disclosure as unlinkability. |
| BBS-derived proofs | Selective disclosure with randomized derived proofs | Can predicates, device performance, holder binding, and revocation meet U-net requirements without introducing correlatable public inputs? |
| AnonCreds | Anonymous credential presentations with predicate and non-revocation mechanisms | Assess witness updates, mobile support, operational complexity, recovery, and independent implementation compatibility. |
| Shared status lists | Fetch status in groups instead of credential-specific status requests | Measure issuer/verifier correlation, stable indexes, caching, freshness, and the size of the actual anonymity set. |
| Private membership/status proofs | Prove current membership against a shared authenticated state | Evaluate witness distribution, update costs, malicious publishers, stale state, and independent cryptographic review. This is research, not an approved custom protocol. |
The baseline specifications are SD-JWT, BBS cryptosuites, AnonCreds, and Bitstring Status List. The table's deployment tradeoffs are questions for U-net's prototype, not claims that these mechanisms already solve U-net's traffic-analysis problem.
Adapters first
Prototype OpenID4VCI issuance and OpenID4VP presentation as adapters around provider-owned storage and wallet consent. Keep transport interoperability separate from the credential format and status mechanism. Preserve per-service account keys, clear consent, retirement, and holder-controlled revocation. Do not label the current Direct Login protocol OAuth/OIDC-compliant.
Required experiment
- Freeze an adversary model covering colluding providers, issuers, gateways, verifiers, messaging operators, and the shared infrastructure administrator.
- Implement two candidates using maintained implementations and synthetic data.
- Capture all public inputs and network metadata over repeated presentations, renewals, revocations, replacements, and persona retirement.
- Measure Android/iOS proof latency, memory, battery cost, offline behavior, and freshness under outages. Measure anonymity sets rather than counting hashes.
- Compare correlation from status lookups, issuer keys, disclosed attributes, delivery routes, timing, and network addresses. Cryptographic unlinkability does not by itself hide traffic patterns.
- Require an independent cryptographic and protocol review, test vectors, and interoperability tests before proposing any production migration.
Deliver a decision record with rejected alternatives, a privacy budget for public inputs, operational costs, and migration/rollback constraints. No production credentials or ledger history are rewritten as part of the experiment.