Microsoft Research Introduces Vega: Fast, Private Credential Verification Without Revealing Identity

Microsoft Research has unveiled Vega, a practical zero-knowledge proof (ZKP) system designed to let people verify facts from government-issued credentials—age, personhood, professional licenses—without ever revealing the credential itself. The system generates cryptographic proofs in under 100 milliseconds on consumer devices, making privacy-preserving identity verification feasible at scale.

The Problem: Identity Verification Requires Trust in Risky Places#

Digital services increasingly need reliable identity signals, especially as AI agents begin transacting on behalf of humans. Yet today's approach is crude: users upload sensitive government IDs to websites and apps, where breaches repeatedly expose them. Regulators are tightening the screws—the EU's digital identity framework and UK age-verification mandates require government ID checks—but the credential-upload model remains the default.

"High-profile breaches have repeatedly exposed government IDs that users shared for routine verification," Microsoft's research team notes. "These are not edge cases."

How Vega Works#

Vega uses zero-knowledge proofs to prove a single fact—say, "I am over 21"—without revealing anything else. The credential never leaves the user's device. The proof works directly on real credential formats like mobile driver's licenses and EU Digital Identity Wallet credentials, requiring no changes from issuing governments.

The system combines several cryptographic innovations:

• Fold-and-reuse proving: Rather than recomputing expensive work for each presentation, Vega precomputes and stores shared credential data once, then reuses it across multiple proofs. This matters for frequent interactions—especially where AI agents present credentials dozens of times daily on a user's behalf.
• Lookup-centric design: Instead of building a full parser into the circuit, Vega treats credentials as byte-addressable tables. The circuit verifies that bytes match the signed credential and pulls specific fields by address—simpler and faster than traditional approaches.
• Length-hiding hashing: SHA-256 hashing, normally expensive in zero-knowledge circuits, is optimized by folding multiple identical step instances into a single proof computation.
• Device binding: The device's private key (stored in secure hardware) signs the proof, ensuring only the phone holder can generate valid proofs for their credential.