VulnFanatic-NG

VulnFanatic-NG LLM 辅助的漏洞研究，用于 Binary Ninja。 VulnFanatic-NG 添加了一个侧边栏面板，扫描当前二进制文件，并询问一个 LLM — 默认是本地托管 的与 OpenAI 兼容的模型，或 Anthropic Claude、Google Gemini、或 Azure OpenAI（见 LLM backends ）— 来判断可疑代码是否真的存在漏洞。它主要基于 Binary Ninja 的反编译器输出（HLIL） ，必要时回退到汇编代码，并且只报告已确认的问题，附带可点击的代码引用。 工作原理 扫描运行最多三个阶段 （第三阶段是可选且仅在线）：...

Investigating Detection and Obfuscation of Prompt Injection Attacks against Software Reverse Engineering AI Agents

Agentic software reverse engineering systems are vulnerable to prompt injection attacks placed into the source code of executable binary files. This research demonstrates defensive tactics for detecting the presences of prompt injection strings in the decompiler output of adversarial example...

ndaybench

ndaybench A benchmark for measuring whether AI agents can bui...