Google DeepMind AI formally verifies Signal encryption code
Google DeepMind announced on November 10 that its AlphaProof Code framework mathematically verified the libsignal library, the open-source core behind Signal, WhatsApp, and Messenger. This marks the first formal verification of such a widely used protocol. The achievement aims to bolster privacy guarantees against evolving AI-driven cyber threats.
Key points
- Google DeepMind announced on November 10 that its AI agents formally verified the libsignal library, the open-source encryption core used by Signal, WhatsApp, and Messenger.
- The verification process utilized a new framework called AlphaProof Code, marking the first time such a widely used protocol has been formally verified by any method.
- The system split the workload between human experts who defined security goals and teams of 40 to 80 AI agents that executed the mathematical proofs.
- Signal CTO Ehren Kret stated the rigorous verification provides greater confidence in privacy guarantees as AI-driven security threats continue to grow.
- The project highlights a shift in formal verification, which traditionally required immense resources, by using AI to automate complex theorem proving for machine code.
AI Verification Milestone for Encrypted Messaging
Google DeepMind has announced a significant breakthrough in software security, stating that its artificial intelligence agents have mathematically proved the integrity of the code behind Signal’s encryption. The announcement was made on November 10 alongside the introduction of a new framework named AlphaProof Code and a detailed technical paper outlining the methodology. This development represents a historic first in the field of cybersecurity, as no previous method, whether AI-driven or traditional, had formally verified the code of such a widely utilized communication protocol.
The core of this achievement lies in the verification of libsignal, the open-source library that powers the encryption for the Signal app, as well as major platforms like WhatsApp, Facebook Messenger, and Google Messages. DeepMind researchers examined the code down to the machine level that runs directly on user devices. By doing so, they ensured that the software adheres strictly to its security promises. This level of scrutiny is critical because libsignal serves as the foundational security layer for billions of users worldwide, making any potential vulnerabilities a matter of global concern.
Formal verification is a rigorous process that uses mathematics to prove that software behaves exactly as intended. Unlike traditional testing, which might find specific bugs, formal verification rules out entire classes of errors by proving logical correctness. Historically, this process has been slow and prohibitively expensive. The technical paper released by DeepMind notes that verifying the seL4 microkernel, a highly secure operating system kernel, required 20 person-years of work. The sheer complexity of verifying a protocol as extensive as libsignal made it a daunting challenge for conventional methods.
To overcome these barriers, DeepMind developed AlphaProof Code, which splits the verification job between human experts and AI agents. Human researchers selected the specific security properties to be proven, while the AI agents handled the heavy lifting of turning these goals into formal specifications. The process involved teams of 40 to 80 AI agents working in parallel to execute the theorem proving. This hybrid approach significantly reduced the time and resources required, demonstrating that AI can accelerate complex mathematical proofs that were previously considered too costly to perform on large-scale software.
The implications of this verification extend beyond just Signal. Because libsignal is the engine behind WhatsApp and Messenger, the mathematical proof of its correctness indirectly bolsters the security claims of these other platforms. However, the verification specifically targeted the libsignal library itself, ensuring that the core cryptographic operations are free from logical flaws. This distinction is important for understanding the scope of the achievement. It does not verify the entire applications, but rather the critical encryption component that handles data privacy.
Ehren Kret, Signal’s chief technology officer, emphasized the importance of this development in the context of modern threats. In a statement published by DeepMind, Kret noted that as AI-driven security threats grow, having the core Signal protocol rigorously verified with automated theorem proving gives the organization even greater confidence that their code delivers on its privacy guarantees. This sentiment reflects a broader industry shift towards leveraging AI not just for attacks, but for defense and validation of critical infrastructure.
The technical paper accompanying the announcement details the method and its limits, providing transparency to the security community. It outlines how the AI agents were trained and how the proofs were constructed. This openness is crucial for independent experts to review and validate the findings. The paper serves as a reference for future projects that may wish to apply similar AI-driven verification techniques to other complex software systems.
Key takeaways
- Google DeepMind used its AlphaProof Code framework to mathematically verify the libsignal encryption library on November 10.
- This is the first formal verification of a widely used messaging protocol by any method, including traditional and AI approaches.
- The process involved human experts defining goals and teams of 40 to 80 AI agents executing the mathematical proofs.
- The verification covers the core code for Signal, WhatsApp, Messenger, and Google Messages, enhancing trust in their privacy guarantees.
- Signal CTO Ehren Kret highlighted the verification as a critical step against evolving AI-driven security threats.
- The project demonstrates that AI can make formal verification, previously too costly, feasible for large-scale software systems.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.