Google DeepMind 使用 AI 助理,為 Signal 的程式碼證明關鍵安全性質,驗證範圍涵蓋編譯後的機器碼。
- - 方法與規模:由 5 人團隊在 4 週內完成。
- 由 Gemini 4 Carbon 驅動的AI 助理,針對未修改的 libsignal 程式碼撰寫約 650 萬行 Lean 證明。
AI 伴隨問答
針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:
這項工作呈現 AI 助理如何與人工撰寫規格、Lean 證明檢查器協作,將 Signal 程式碼的安全性驗證延伸至編譯後機器碼。
Banger report from Google DeepMind.
This is also a good example of the type of ambitious work you can now do with frontier models.
(bookmark it)
They used AI agents to prove key security properties of Signal's code, down to the compiled machine code.
Agents powered by Gemini 4 Carbon wrote about 6.5 million lines of Lean proofs on the unmodified libsignal code. A team of five finished the proof work in four weeks. Verifying the seL4 microkernel took twenty person-years.
In this work, humans and agents write the security specs, prover agents write the proofs, and Lean checks every proof before it is accepted.
When they reintroduced two old Signal bugs, the system proved the buggy versions were insecure.
Paper: https://t.co/aKJlFHMn5y