Janus:面向 Android 原生库的 Agent 跨语言模糊测试 harness 生成
原标题:JANUS: Agentic Cross-Language Fuzzing Harness Generation for Android Native Libraries
AI 摘要
南开大学等机构的研究者提出 Janus,一种基于 LLM Agent 的方法,用于为 Android 原生库自动生成跨语言 JNI 感知的模糊测试 harness。该方法通过迭代解析显式和隐式的跨语言调用依赖、自底向上递归构造跨语言参数对象,减少对手工启发式规则的依赖。在 44 个流行真实应用(共 21,386 个 JNI 方法)上,Janus 生成的有效 harness 数量比现有最优工具多 5.9 倍以上(6,094 个),并在五家主流 Android 厂商的预装应用中发现了 6 个此前未知的闭源原生库漏洞。
正文节选
Janus: Agentic Cross-Language Fuzzing Harness Generation for Android Native LibrariesThanks: 1 Lingling Fan is the corresponding author (linglingfan@nankai.edu.cn). Abstract Modern Android apps extensively rely on native libraries, bridging Java/Kotlin and C/C++ through the Java Native Interface (JNI). This cross-language interoperation, however, exposes the memory-safe Android Runtime to the unsafe C/C++ world, creating an attack surface that has been repeatedly exploited for full device compro