VOLE-Based Zero-Knowledge
A zero-knowledge proof lets a person prove a fact about their identity without revealing the underlying data. Generating one used to be slow and memory-hungry, which pushed the work onto servers and away from the person’s own device. human.tech uses a VOLE-based proving system so the proof is generated on the person’s own hardware, in the browser, in seconds.
Why the proving system matters
Where a proof is generated decides who holds the data while it is generated. If proving happens on a server, the person’s raw document, biometrics, and personal data have to reach that server. Generating the proof on the person’s own device keeps that data on the device. The constraint has always been performance: proving a passport signature or a large hash inside a general-purpose zero-knowledge system is expensive.
VOLE-based proving removes that constraint. It “facilitates rapid identity proofs while maintaining privacy on consumer hardware” and lets “mobile browsers complete passport proofs within seconds without memory constraints.”
How it works
The system uses VOLE (Vector Oblivious Linear Evaluation) “in the head,” with a modified linear code based on the repeat-multiple-accumulate (RMA) code. At its core is a two-party commitment: the verifier obtains functions describing collinear points, and the prover receives random points along those lines.
Proving then has two phases:
- Commitment. A derandomization step turns random-value commitments into commitments to the actual witness (the private inputs).
- Proving the circuit. Addition gates resolve trivially because the commitment is linearly homomorphic. Multiplication gates are proven with committed values and correctness proofs using Quicksilver.
The system integrates existing, audited circom circuit implementations for the heavy operations that identity proofs depend on, such as RSA, SHA256, and Keccak256, rather than reimplementing them.
The trade-off
VOLE-based proving is deliberately optimized for the identity case: proofs that a person generates for a verifier, fast, on their own device. It “achieves dramatically better efficiency than mainstream proving systems, though sacrifices the succinctness needed for rollups.” Making the proof non-interactive (“in the head”) adds roughly 2x overhead. The result is a proof that is quick to produce on a phone rather than one small enough to post to a rollup, which is the right trade for identity.