LLMs Break Cryptographic Schemes in New CryptanalysisBench Study

AI 中文解读

AI竟然会破解密码了！最新研究显示，顶尖大模型能成功破解65%到86%的有漏洞加密方案，甚至发现了以前从未被攻破的密码算法的新漏洞。这就像给AI装上了“密码破译大脑”，它不再只是聊天写诗，而是能像顶级数学家一样，在密码系统里找到隐藏的弱点。比如Anthropic的Mythos模型，就自己发现了“鹰牌签名算法”和简化版的AES加密的漏洞，这是人类首次看到AI独立完成这种密码分析工作。这对普通人意味着什么呢？你平时用手机支付、发微信、登录网站，背后都靠密码算法保护数据。如果AI破解密码的能力超过人类专家，现有的加密体系可能面临威胁，未来我们得用更复杂的密码来保护隐私。不过也不用太担心，这项技术也能帮助安全专家提前发现漏洞，升级防护措施。简单说，AI既能“拆门”，也能帮我们“修门”——关键在于谁能跑得更快。

<h3> Forensic Summary </h3> <p>A new benchmark, CryptanalysisBench, demonstrates that frontier LLMs can perform meaningful cryptanalysis, breaking 65–86% of schemes with known practical vulnerabilities and producing novel attacks against previously unbroken primitives. Anthropic's Mythos Preview model uncovered new vulnerabilities in the Hawk signature scheme and reduced-round AES, representing the first AI-discovered cryptanalytic results of this kind. This signals a near-term shift in the threat landscape where AI-assisted cryptanalysis may begin to outpace human expert analysis.</p> <p>Read the full technical deep-dive on <strong>Grid the Grey</strong>: <a href="https://gridthegrey.com/posts/llms-break-cryptographic-schemes-in-new-cryptanalysisbench-study/" rel="noopener noreferrer">https://gridthegrey.com/posts/llms-break-cryptographic-schemes-in-new-cryptanalysisbench-study/</a> </p>