Anonymous proof of humanity with zero-knowledge proofs
Feasibility assessment and a working proof-of-concept prototype, with the protocol split into an MVP core (attestation, risk engine, partner API) and a ZK phase (proofs, Ethereum anchoring, the top trust levels).
Stack and topics
| Industry | Web3 identity and anti-fraud |
|---|---|
| Region | Global |
| Scale | Six trust levels, L0 to L5 |
Context
The client and the problem
Airdrops, DAO votes and reward programmes were being drained by bots and multi-accounts. The usual fixes ask for a passport or a face scan, which many users refuse and which leaves a store of personal data that someone has to protect.
A protocol team wanted to prove that one person is one participant without KYC, without biometrics and without storing personal data, with trust levels that partners can rely on.
What we built
What did we build?
We assessed whether such a protocol is feasible, designed it in layers and built a proof-of-concept prototype.
- Device attestation. A challenge-response nonce checked through Secure Enclave and DeviceCheck on iOS and Play Integrity and Keystore on Android, with root, jailbreak and emulator detection and honeypot signals.
- Behavioural risk engine. An event collector, feature extraction, an ML risk score, a rotating anonymous profile identifier and a global risk graph that links devices and patterns, never names.
- Zero-knowledge layer. The user proves "my score is above the threshold for this level" without revealing any input. Groth16 circuits come first for cheap verification on Ethereum, with a path to Halo2 and recursive proofs; proofs are generated on the client or through a relayer and verified in a smart contract.
- Trust levels on chain. Six levels from L0 to L5. Everyday levels live on a low-cost chain; high-trust levels are anchored on Ethereum mainnet with staking, slashing and soulbound-style status records.
Protocol
How does the proof work?
The risk engine sends an attestation challenge with a nonce, and the device answers with an attestation token and behaviour events. The engine returns a commitment to the risk score. On the device, a Groth16 proof shows that the score is at or above the threshold for the requested level, without revealing the score or any input. A smart contract verifies the proof, and partner applications read the verified level on-chain.
How we worked
Team and process
The work started with a feasibility assessment: which signals are strong enough, what can be proven in zero knowledge at a cost users accept, and where the attacks are. Only then did we build the prototype.
We split the protocol into an MVP core that a team can ship first (attestation, risk engine, partner API) and a ZK phase (proofs, Ethereum anchoring and the top trust levels), so the client can launch without waiting for the hardest part.
Facts
Project facts
| Fact | Detail |
|---|---|
| Client | A Web3 identity and anti-fraud protocol team |
| Region | Global |
| Scope | Feasibility assessment and proof-of-concept prototype |
| Trust levels | Six, L0 to L5 |
| Stack | Groth16, Halo2 path, Secure Enclave, DeviceCheck, Play Integrity, Ethereum, TON |
Result
Where it stands
The client has a feasibility assessment and a working proof-of-concept prototype, with the protocol split into an MVP core and a ZK phase. This is a study and a prototype, not a production launch.
Next step
Building something similar?
A 30 minute call with Oleg. You leave with a clear view of scope, risks and the next step.
Oleg Pun
CEO, Vareger
