Google DeepMind melds humans and AI agents to vet Signal's secure messaging techniques

Google DeepMind researchers have successfully completed a formal verification of Signal's messaging security using AlphaProof Code, a multi-agent framework for formal software specification and verification. This marks the first time AI has helped devise a proof linking protocol-level security guarantees to a widely used cryptographic implementation.

The verification process involved creating a human-reviewed mathematical specification, generating Lean code to prove the correctness of libsignal implementation, and performing an auditing pass that required human review. The researchers claim this achievement is a step towards the challenge issued by the Beneficial AI Foundation in April to formally verify the Signal protocol and app with Lean.

While the theoretical nature of the result assumes perfect cryptography, it still represents a significant assertion of correctness, demonstrating the potential of human-AI collaboration in formal methods.

Brief written by urgent.news from The Register Science's own syndicated text. Machine-written — may contain errors; check the original before relying on it.