DeepMind formally verifies Signal protocol guarantees with AI
Original title: Formal Verification of the Signal Protocol Using AI
In a four-week proof of concept, Gemini Carbon agents produced over six million lines of Lean and re-found only already-fixed bugs.
Google DeepMind researchers used a framework called AlphaProof Code to formally verify key cryptographic guarantees of the Signal protocol. Experts reviewed translated security specifications, then Gemini Carbon agents generated the proofs and flagged about one percent of the Lean codebase for human audit. The analysis found no new vulnerabilities in the portions examined, while older Signal versions with known bugs failed verification. The researchers are releasing the specification and proof.
Unconfirmed: The finding of no new vulnerabilities is the DeepMind researchers' own result; independent review is not described.
- The work covered cryptography used by Signal, Google Messages, and WhatsApp.
- Human experts reviewed every formal security specification before proofs began.
- Agents marked about one percent of the codebase for a second human look.
- On older Signal versions, the system identified known, long-fixed vulnerabilities.
- Vendors
- Google DeepMind
- AI models
- Gemini Carbon
From the original article
1 min read at schneier.comThis seems like a good thing: Google Deepmind researchers used agentic AI to mathematically prove, or “formally verify,” a number of the Signal protocol’s key cryptographic guaranteesa landmark development for the widely used cryptography and AI itself. Conducted as a sort of proof-of-concept project in just four weeks, the work is significant in two ways.
By Bruce Schneier · schneier.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.