arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2410.03470cs.LGcs.AImath.AT
收藏

基于注意力图拓扑分析的漏洞检测

Vulnerability Detection via Topological Analysis of Attention Maps

Institute for Information Transmission Problems of Russian Academy of Sciences(俄罗斯科学院信息传输问题研究所)

Pavel Snopov, Andrey Nikolaevich Golubinskiy

AI总结:

本文提出利用拓扑数据分析工具从BERT注意力矩阵中提取特征,使传统机器学习模型在漏洞检测上达到与预训练语言模型CodeBERTa相当的性能。

摘要:

近年来,基于深度学习(DL)的漏洞检测方法获得了显著关注。这些方法展示了有前景的结果,在有效性上往往超越了传统的静态代码分析工具。在本研究中,我们探索了一种新颖的漏洞检测方法,该方法利用拓扑数据分析(TDA)工具对BERT模型的注意力矩阵进行分析。我们的发现表明,传统的机器学习(ML)技术在使用从这些注意力矩阵中提取的拓扑特征进行训练后,能够与预训练语言模型(LLMs)如CodeBERTa相竞争。这表明包括持续同调在内的TDA工具能够有效捕获对识别漏洞至关重要的语义信息。

英文摘要:

Recently, deep learning (DL) approaches to vulnerability detection have gained significant traction. These methods demonstrate promising results, often surpassing traditional static code analysis tools in effectiveness. In this study, we explore a novel approach to vulnerability detection utilizing the tools from topological data analysis (TDA) on the attention matrices of the BERT model. Our findings reveal that traditional machine learning (ML) techniques, when trained on the topological features extracted from these attention matrices, can perform competitively with pre-trained language models (LLMs) such as CodeBERTa. This suggests that TDA tools, including persistent homology, are capable of effectively capturing semantic information critical for identifying vulnerabilities.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑