Google DeepMind researchers say their AI agents have produced mathematical proofs linking specific security guarantees to a compiled build of Signal's encryption library, with results published online for public checking. However, the audit's scope excludes key parts of the messaging app, including key generation and group messaging.
The Nov. 10 AlphaProof Code report describes about 6.5 million lines of machine-checked Lean proof code. The result concerns a particular build of libsignal, the Signal protocol's implementation, under explicit assumptions about cryptography and the environment in which the code runs. The authors say five people worked on the proofs over four weeks. Humans directed the work, selected reference models, and reviewed the specifications that defined what the agents had to prove.
The target was an AArch64 Linux executable using libsignal-protocol v0.105.4 and SPQR v1.6.0 inside a small command-line harness. The theorems cover session initialization, sending a message, and receiving one. Key generation falls outside that verified scope.
The Google report connects the compiled executable to specified properties of PQXDH and Triple Ratchet one-to-one messaging, including forward secrecy, authentication, and post-compromise security. These address whether earlier messages stay secret after keys are stolen, whether senders are authentic, and whether a session can regain secrecy after a compromise.
Its two proof models provide different assurances: the symbolic analysis relies on idealized cryptographic operations and collision-free executions, while the computational analysis bounds the success of classical attackers under stated cryptographic assumptions. The symbolic proof establishes handshake secrecy under specified later breaks of Diffie-Hellman, but the computational proof does not cover quantum attackers. Group messaging is outside the report's scope, and the computational model excludes multi-device state. Its bounds also stop providing a useful guarantee at sufficiently large workloads.
The results still require trust in Lean's checking kernel, the specifications reviewed by the researchers, the AArch64 Linux model, and contracts describing selected routines. The project's Nov. 11 update, merged on GitHub, says ten more library routines were verified, cutting trusted routine hooks from 37 to 27. Those hooks replace execution of selected routines with assumed descriptions of their behavior. The update retains libsignal v0.105.4 but changes the pinned executable. The newer build verifies the machine code for those ten routines instead of relying on their modeled contracts.
The project repository provides public instructions to rebuild the pinned binary and check its proofs. A Linux-only comparator checks that the theorems and their dependent definitions match trusted statements, then replays proofs in a fresh Lean kernel. The authors also report reintroducing two previously fixed libsignal bugs and proving that the modified binaries violate their symbolic security theorem. That tests whether the specification rejects those known regressions.
Analysis
Why This Matters
- The verification approach could signal a new paradigm for auditing critical security software, using AI-generated proofs that are independently checkable.
- The gaps in verification, key generation, group messaging, quantum attacker resistance, mean the widely used Signal app retains unverified vulnerabilities that could be exploited in practice.
- The project's public reproducibility lets other researchers and security professionals independently verify the claims, increasing transparency.
Background
Google DeepMind developed AlphaProof, an AI agent that writes machine-checked proofs in the Lean theorem prover. The project aims to formally verify software security properties by linking compiled executables to mathematical specifications. The Signal Protocol, used by Signal and other messaging apps, is a high-profile target due to its widespread adoption for encrypted communication.
Key Perspectives
Google DeepMind researchers: attribute to the authors' report, which says their AI agents produced proofs linking specific security guarantees to a compiled build of Signal's encryption library, and they put the resulting evidence online for others to check.
What to Watch
- Whether other researchers independently check and confirm the proofs using the public repository.
- Whether the gap in key generation and group messaging verification is addressed in future updates.
- How the project handles the trusted computing base, as it currently relies on Lean's kernel, specifications, and the AArch64 machine model.
Checked
12 claims in this story checked against 1 source: 12 supported.
Show each claim
- SupportedGoogle AI agents audited Signal code, with one major flaw staying hidden.CryptoSlate: “Google AI agents audit Signal’s code and one major flaw stays hidden.”
- SupportedDeepMind researchers say AI agents produced mathematical proofs linking security guarantees to a compiled build of Signal’s encryption library.CryptoSlate: “Google DeepMind researchers say their AI agents produced mathematical proofs linking specific security guarantees to a compiled build of Signal’s encryption library”
- SupportedThe resulting evidence was put online for others to check.CryptoSlate: “and they put the resulting evidence online for others to check.”
- SupportedThe audit scope excludes key generation and group messaging.CryptoSlate: “No direct proof of the phone apps, key generation or group messaging.”
- SupportedThe Nov. 10 report describes about 6.5 million lines of machine-checked Lean proof code.CryptoSlate: “The Nov. 10 AlphaProof Code report describes about 6.5 million lines of machine-checked Lean proof code.”
- SupportedFive people worked on the proofs over four weeks.CryptoSlate: “The authors say five people worked on the proofs over four weeks.”
- SupportedThe target was an AArch64 Linux executable using libsignal-protocol v0.105.4 and SPQR v1.6.0 in a small command-line harness.CryptoSlate: “The target was an AArch64 Linux executable using libsignal-protocol v0.105.4 and SPQR v1.6.0 inside a small command-line harness.”
- SupportedThe theorems cover session initialization, sending a message and receiving one, but not key generation.CryptoSlate: “The theorems cover session initialization, sending a message, and receiving one. Key generation falls outside that verified scope.”
- SupportedThe symbolic proof establishes handshake secrecy under specified later breaks of Diffie-Hellman, while the computational proof does not cover quantum attackers.CryptoSlate: “The symbolic proof establishes handshake secrecy under specified later breaks of Diffie-Hellman, but the computational proof does not cover quantum attackers.”
- SupportedThe Nov. 11 update says ten more library routines were verified, cutting trusted routine hooks from 37 to 27.CryptoSlate: “The project’s Nov. 11 update, merged on GitHub, says ten more library routines were verified, cutting trusted routine hooks from 37 to 27.”
- SupportedThe newer build verifies the machine code for those ten routines instead of relying on their modeled contracts.CryptoSlate: “The newer build verifies the machine code for those ten routines instead of relying on their modeled contracts.”
- SupportedThe authors report reintroducing two previously fixed libsignal bugs and proving that the modified binaries violate their symbolic security theorem.CryptoSlate: “The authors also report reintroducing two previously fixed libsignal bugs and proving that the modified binaries violate their symbolic security theorem.”