Skip to content
Technology

Google Uses AI to Verify Signal Message Security

Researchers from Google DeepMind successfully employed artificial intelligence to provide formal security verification for the messaging application Signal.

By 2 min read

Researchers at Google DeepMind have successfully utilized artificial intelligence to perform formal security verification for messages sent through the Signal application. This achievement marks the first time that automated systems have helped create a mathematical proof linking the security guarantees of a protocol to its actual code implementation, according to a technical report recently released.

The initiative leveraged AlphaProof Code, an AI agent-based system designed for software specification and verification. Through this technology, the team validated that the application protects conversations and prevents data theft, successfully addressing a challenge originally issued by the Beneficial AI Foundation earlier this year.

The procedure involved transforming security goals into mathematical specifications that were reviewed by humans, while the AI model generated millions of lines of code in Lean, a specialized programming language, to test software correctness. Swat Chaudhuri, a researcher at Google DeepMind and professor at the University of Texas in the United States, highlighted the significance of this human-machine collaboration to the specialized publication The Register.

"In formal verification, we always had the belief that verification is the difficult part. There are 30 or 40 years of work on how to scale verification. Now, we managed to do that. So, now the question is: what are we verifying?" Chaudhuri stated.

As quantum computers advance and present potential threats to traditional cryptography based on elliptic curves, the Signal platform has implemented defenses to safeguard current and future data. The introduction of the Sparse Post Quantum Ratchet, combined with the existing system, forms the Triple Ratchet, which blends traditional and post-quantum keys through derivation functions to establish a reinforced protective framework.

The verification of such complex protocols traditionally demanded extensive manual effort, such as the years invested in previous systems. By utilizing AlphaProof Code, the team reduced operational timelines and expenses while maintaining continuous integration within the software development workflow on GitHub.

Looking forward, Chaudhuri expressed an ambition to scale the technology further, pointing out that the next objective involves bringing this automated verification directly into the design phase for new tools across digital systems.

Sources

Related stories