二进制代码语义恢复技术研究综述OA
A Survey on Binary Code Semantic Recovery Technology
二进制代码语义恢复技术旨在在缺失源代码与调试信息的条件下,重建函数、变量、数据结构等高层语义信息,以支撑恶意代码理解、漏洞定位与软件成分识别.受符号剥离、编译优化与跨平台差异影响,高层语义与低层实现之间存在信息缺失与形态扰动并存的问题,导致信息稀疏、表示异构及全局依赖难以刻画.系统回顾代表性研究,将现有方法归纳为启发式规则匹配、深度学习判别式推断与大语言模型驱动的生成式恢复3类范式,并梳理其演进脉络,进一步总结主流数据集与评测指标的发展趋势.最后,剖析了该领域研究面临的真实挑战,并展望了未来研究方向.
Binary code semantic recovery technology aims to reconstruct high-level semantic informa-tion such as functions,variables,and data structures in the absence of source code and debugging in-formation,to support malicious code understanding,vulnerability localization,and software component identification.Due to symbol stripping,compiler optimizations,and cross-platform differences,there coexist problems of information loss and morphological perturbation between high-level semantics and low-level implementations,leading to sparse information,heterogeneous representations,and difficul-ties in characterizing global dependencies.Representative studies are systematically reviewed in this paper.The existing methods are categorized into 3 paradigms,which include heuristic rule matching,deep learning discriminative inference,and large language model-driven generative recovery,and their evolution is traced.The development trends of mainstream datasets and evaluation metrics are further summarized.Finally,the real challenges faced by research in this field are analyzed and future re-search directions are anticipated.
孙佳佳;李泽潭;付艳华;朱肖雅;尹小康;杨启超;刘胜利
信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001信息工程大学,河南 郑州 450001
信息技术与安全科学
二进制代码语义恢复逆向工程代码理解深度学习大语言模型
binary codesemantic recoveryreverse engineeringcode comprehensiondeep learn-inglarge language model
《信息工程大学学报》 2026 (4)
443-449,7
评论