Privacy-preserving verification

01 research

The question

Verification and privacy are usually purchased with each other's currency: to prove you are unique, you disclose who you are; to prove a computation happened, you reveal its inputs. Zero-knowledge proofs, selective disclosure, and hardware attestation each attack the trade-off from a different angle. The research question is where the composition of the three is practical today at infrastructure scale — and where it still is not.

02 research

Identity: prove the property, not the person

The VEID design aims to verify selected identity attributes without exposing source documents. Original document images and full OCR output remain local and are not upload payloads; a user may approve submission of minimum derived data for a named purpose. Repository modules and design documents describe intended capabilities and do not establish that every proof, device-attestation, or selective-disclosure flow is deployed.

03 research

Computation: verifiable records without exposed internals

DSEMA applies the same discipline to machine behaviour. Every agent action is signed against its DID; structured performance records are appended to the chain with roll-up proofs; full logs are published to content-addressed storage (IPFS) with only CIDs and hashes on-chain; and verifiable credentials attest to specializations without exposing model internals. The result under study is an audit trail that is complete enough for accountability and sparse enough for confidentiality.

Constitutionally, clause 39.2 requires preferring architectures that use encryption and selective disclosure where feasible and allow independent audit where safe and lawful — making this stream an obligation, not an aesthetic.