The Last Mile of Attack Investigation: Audit Log Analysis Toward Software Vulnerability Location
Changhua Chen,Tingzhen Yan,Chenxuan Shi,Hao Xi,Zhirui Fan,Hai Wan,Xibin Zhao
DOI: https://doi.org/10.1109/tifs.2024.3459616
IF: 7.231
2024-10-18
IEEE Transactions on Information Forensics and Security
Abstract:Cyberattacks have caused significant damage and losses in various domains. While existing attack investigations against cyberattacks focus on identifying compromised system entities and reconstructing attack stories, there is a lack of information that security analysts can use to locate software vulnerabilities and thus fix them. In this paper, we present AiVl, a novel software vulnerability location method to push the attack investigation further. AiVl relies on logs collected by the default built-in system auditing tool and program binaries within the system. Given a sequence of malicious log entries obtained through traditional attack investigations, AiVl can identify the functions responsible for generating these logs and trace the corresponding function call paths, namely the location of vulnerabilities in the source code. To achieve this, AiVl proposes an accurate, concise, and complete specific-domain program modeling that constructs all system call flows by static-dynamic techniques from the binary, and develops effective matching-based algorithms between the log sequences and program models. To evaluate the effectiveness of AiVl, we conduct experiments on 18 real-world attack scenarios and an APT, covering comprehensive categories of vulnerabilities and program execution classes. The results show that compared to actual vulnerability remediation reports, AiVl achieves a 100% precision and an average recall of 90%. Besides, the runtime overhead is reasonable, averaging at 7%.
computer science, theory & methods,engineering, electrical & electronic