A sovereign verification protocol using zero-knowledge proofs and a tamper-evident registry, so a verifier learns one fact instead of receiving an identity document.
A bar needs to know you are over 18. A bank needs to know you live where you say you do. In both cases the verifier asks one question and receives an entire identity document: name, address, date of birth, document number.
That creates three costs at once. Manual verification is slow and expensive, calls and emails and visual checks against siloed databases. Human checking introduces error and social-engineering risk. And every copy stored becomes a honeypot, which is a liability under GDPR and CCPA long before anyone breaches it.
So the design target was not better fraud detection. It was to make the verifier receive less.
The government issues a cryptographically signed credential once. The holder’s wallet generates a zero-knowledge proof locally, on the device, that a specific statement is true. The verifier gets a boolean and a validity check.
Raw data never leaves the phone. The issuer database is not called on every check, which removes both the latency and the availability dependency. The verifier stores no personal data, so there is nothing to leak and nothing to be compliant about.
Everything sensitive is off-chain: the government PII database, the issuer service, the holder wallet, the verifier SDK. The chain holds four registries and nothing else. Issuer keys mapped to DIDs. Schema and circuit hashes, which whitelist the proof circuits a verifier will accept. Revocation status as bitmaps and Merkle roots per epoch, giving constant-time status checks. Audit checkpoints.
The chain is a public bulletin board for keys, schemas and revocation roots. It never sees a person. That boundary is the design, and keeping the on-chain surface minimal is also what keeps the attack surface and the gas costs down.
A permissioned committee of five to fifteen known validators running BFT or proof of authority gives instant deterministic finality, no gas, no reorgs, and a validator set a ministry can govern. The trade-off is explicit: it requires trusting the domestic authority, which is less credible to an international verifier.
So state roots are periodically checkpointed to an Ethereum L2. That buys global integrity and tamper-evidence, at the cost of slower finality and variable gas, offset by batching.
The honest framing is the one on the slide: this is not trustless. It assumes at least two thirds of validators are honest, issuer keys are secure, and the anchor publisher stays live. Sovereign finality is sufficient for domestic use. Anchoring is what makes a foreign verifier able to check the history independently.
Verification happens at the edge between wallet and verifier SDK, so there is no central server to bottleneck and throughput is effectively unlimited. Latency is client-side proof generation, roughly one to two seconds, not network congestion.
Writes are batched: thousands of issuances and revocations compress into a single Merkle root update, one transaction per epoch. The chain load is independent of how many verifications happen.
Five attack vectors with a defence each, because assuming components fail is cheaper than assuming they do not.
A stolen issuer key signs fake credentials, so keys live in hardware security modules with instant revocation registry rotation. A malicious contract upgrade is contained by a 48-hour timelock plus an emergency pause. Revocation manipulation is caught by rate limits on status updates and multisig consensus. Validator censorship is handled by committee diversity and a cached verification fallback so verifiers degrade rather than stop. A flawed ZK circuit that lets false proofs pass is the worst case, which is why circuits are versioned and auditable and unsafe versions can be deprecated.
Governments only, deliberately. The customers are border and immigration, interior and home affairs, transport authorities, and health and welfare entitlements. Industry estimates put the government digital identity segment above $15bn with roughly 18% CAGR.
Two revenue lines: a micro-fee paid by the government per validity check, priced like a credit bureau lookup, and an enterprise SDK subscription for the verifiers who consume it, banks and airports.
No token. That is a commercial decision rather than a technical one: a token invites speculation and makes the product unbuyable through government procurement.
The competitive gap is specific. Traditional identity verification, Onfido and Jumio, relies on document photos and biometrics per check, which is exactly the high-friction, high-PII-risk model this replaces. Government integrators like IDEMIA and Thales have the relationships and the centralised silos, but not the privacy-preserving cryptography.
Rollout is staged: one claim, age over 18, with one government partner, then the full credential suite, then cross-border verification through L2 anchoring.