一句话：提出 NativeSummary，通过提取原生二进制语义摘要并转换 JNI/原生调用，实现高精度、可扩展的 Android 跨语言静态分析。
Android 静态分析长期以 DEX 字节码为主要目标，而现代应用中普遍存在的 C/C++ 原生代码常被忽略；少数处理原生代码的工作仍受可扩展性与精度限制。本文提出 NativeSummary：先提取原生二进制代码的语义摘要，再将 JNI 接口函数的常见使用模式转换为 Java 字节码操作，并把原生库函数调用转换为字节码调用。这样无需改动现有 Java 静态框架的固有逻辑，即可为其赋予跨语言数据流分析能力。大量评估表明，NativeSummary 在精度、可扩展性与兼容性上优于现有最优技术，并使既有应用分析工作几乎无需额外成本即可受益。
With the prosperity of Android app research in the last decade, many static analysis techniques have been proposed. They generally aim to tackle DEX bytecode in Android apps. Beyond DEX bytecode, native code (usually written in C/C++) is prevalent in modern Android apps, whose analysis is usually overlooked by most existing analysis frameworks. Although a few recent works attempted to handle native code, they suffer from scalability and accuracy issues. In this paper, we propose NativeSummary, a novel inter-language static analysis framework for Android apps with high accuracy, scalability, and compatibility. Our key idea is to extract semantic summary of the native binary code, then convert common usage patterns of JNI interface functions into Java bytecode operations, and additionally transform native library function calls to bytecode calls. Along with this effort, we can empower the legacy Java static frameworks with the ability of inter-language data flow analysis without tampering their inherent logic. Extensive evaluation suggests that NativeSummary outperforms SOTA techniques in terms of accuracy, scalability and compatibility. NativeSummary sheds light on the promising direction of inter-language analysis, and thousands of existing app analysis works can be boosted atop NativeSummary with almost no effort.
DOI 原文 ·
@inproceedings{paperbot8537,
title = {NativeSummary: Summarizing Native Binary Code for Inter-language Static Analysis of Android Apps},
author = {Jikai Wang and Haoyu Wang},
booktitle = {ISSTA},
year = {2024},
doi = {10.1145/3650212.3680335}
}