Google DeepMind 用 AlphaProof Code 验证 Signal 协议实际代码的关键安全属性Chinese only
Google DeepMind 研究人员让 AI 智能体与人类专家协作，对 Signal 协议的实际代码进行形式化验证，证明其符合指定的关键安全属性。研究团队称，验证耗时四周；他们还将两处已修复漏洞重新加入隔离版本，系统均给出了机器可检查的“不安全”证明。该方法的扩大应用仍需研究。
1 developments5 reports3 sources4 verified facts
Verified facts 4
Introducing AlphaProof Code: a breakthrough in verifying complex software at depth
Gemini Carbon 驱动的智能体完成了 Signal 协议代码的形式化证明，并生成了超过六百万行 Lean 代码。
谷歌 DeepMind 用 AI 完成 Signal 协议形式化验证，数学证明其端到端加密可靠性
谷歌 DeepMind 团队利用人工智能对 Signal 加密协议的安全性进行了形式化验证，并用该方法识别出旧版协议实现中的已知安全问题。
Event developments 1
Google DeepMind uses AlphaProof Code to verify key security properties of Signal's implementation
Google DeepMind researchers combined AI agents with human experts to formally verify that Signal's implementation meets specified key security properties. The team says the verification took four weeks. After reintroducing two previously fixed bugs into an isolated version, the system produced machine-checked proofs that both buggy versions were insecure. Scaling the approach requires further research.
View source reports · 3 sources · 5 reports
- Google DeepMind 官方博客 · Google DeepMind uses AlphaProof Code to verify key security properties of Signal's implementation
- IT之家 科技新闻 · Google DeepMind uses AI to formally verify the Signal encryption protocol
- Rohan Paul (@rohanpaul_ai) · Google DeepMind Uses AI to Verify Key Security Properties of Signal’s Production Code
- Rohan Paul (@rohanpaul_ai) · Google DeepMind says AI verified key security properties of Signal’s machine code
- Rohan Paul (@rohanpaul_ai) · Post says Google DeepMind used AI agents to verify secrecy in Signal’s running code
All reports 5
Post says Google DeepMind used AI agents to verify secrecy in Signal’s running code
A post describing the work says AI agents wrote about 6.5 million lines of Lean proofs to check message secrecy in Signal’s compiled code, with a strict proof checker rejecting incorrect steps. It also says researchers reintroduced two previously fixed bugs, and the system proved the code insecure both times.
3 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- IT之家 科技新闻
- Google DeepMind 官方博客
Google DeepMind says AI verified key security properties of Signal’s machine code
Google DeepMind says AlphaProof Code used models including Gemini Carbon to formally verify code running on devices for the Signal protocol. Humans set the security goals and Lean checked the AI-generated proofs. When researchers reintroduced two previously fixed bugs, the system produced machine-checked proofs that the code was insecure both times. The team is open-sourcing its specifications and verification artifacts.
3 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- IT之家 科技新闻
- Google DeepMind 官方博客
Google DeepMind Uses AI to Verify Key Security Properties of Signal’s Production Code
Google DeepMind’s team says AlphaProof Code helped it verify key security properties of the Signal protocol down to code running on devices in about a month. The team says it is open-sourcing its formal specifications and verification artifacts, while further research is needed.
3 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- IT之家 科技新闻
- Google DeepMind 官方博客
Google DeepMind uses AI to formally verify the Signal encryption protocol
Google DeepMind used AI to formally verify the Signal encryption protocol and used the same method to identify known security flaws in older protocol implementations. The work verified the protocol, not the entire Signal app; the analysis took about four weeks, and Google says it cannot yet determine whether the method is feasible at a larger scale.
2 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- Google DeepMind 官方博客
Google DeepMind uses AlphaProof Code to verify key security properties of Signal's implementation
Google DeepMind researchers combined AI agents with human experts to formally verify that Signal's implementation meets specified key security properties. The team says the verification took four weeks. After reintroducing two previously fixed bugs into an isolated version, the system produced machine-checked proofs that both buggy versions were insecure. Scaling the approach requires further research.
2 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- IT之家 科技新闻