Google DeepMind researchers have used agentic artificial intelligence to mathematically prove, or formally verify, several of the key cryptographic guarantees of the Signal protocol in a proof‑of‑concept project that lasted just four weeks.
The work provides an independent check of the cryptography that underpins the security of millions of messages sent on platforms such as Signal, Google Messages and WhatsApp, giving fresh confidence in the protocols that protect everyday communications.
The verification was carried out with AlphaProof Code, a new AI framework developed by DeepMind that combines machine learning with formal methods to automatically generate proofs that software behaves as intended.
AI‑aided formal verification is gaining attention, but experts note that its practicality for widespread adoption and its effectiveness on large codebases remain uncertain. Proponents see it as a “holy grail” of software security, using mathematics to rule out critical vulnerabilities.
Formal verification works by creating a mathematical model of the software and then proving, with no doubt, that the model satisfies its specifications—an approach that offers a level of assurance beyond traditional testing.
<small>Source: Ars Technica — read the original story there.</small>
