基于指令关系图的二进制代码与源代码相似性分析方法，首发论文 -全讯担保网

binary-source code similarity comparison

• 1、state key laboratory of networking and switching technology, beijing university of posts and telecommunication, beijing 100876, china

abstract：the analysis and reverse engineering of executable binary programs have extensive applications in various fields, including computer security. reverse engineering programs in binary form are often considered a manual and time-consuming process, making it challenging to efficiently apply them to large sets of samples. however, security companies often need to analyze thousands of unknown binary files daily, necessitating fast and automated methods for binary analysis and reverse engineering. therefore, the similarity analysis of binary code to source code plays a crucial role in the analysis and reverse engineering of binary programs. existing methods have limitations in extracting code semantics, as some methods overlook certain semantic information while others include redundant information. hence, this paper proposes a binary-to-source code similarity analysis method based on llvm. this method generates an instruction relationship graph from llvm-ir, obtained from both binary and source code, capturing the control flow, data flow, and call flow relationships between instructions. additionally, a code similarity analysis model based on graph neural networks is introduced, which extracts semantic features from the code using the instruction relationship graph and a globally pooled control flow graph. the proposed method is evaluated on the open-source dataset codenet and its performance is assessed for different compilation options of binary code. experimental results demonstrate that the proposed method outperforms existing state-of-the-art methods.keywords：reverse engineering, clone detection, intermediate representation, binary code, code matching, deep learning.

keywords：