A Novel Method of Software Vulnerability Detection based on Fuzzing Technique

摘要

Buffer overflow vulnerabilities can cause attacks that result in serious consequences. However the techniques of buffer overflow vulnerability detection are limited to manual analysis, binary-patch comparison, fuzzing and so on. They rely on manual analysis, thus cause high overhead. In this paper, we propose a novel method of detection of buffer overflow vulnerabilities, which is based on fuzzing, data-flow dynamic analysis and automated exception analysis. This new method effectively improves the detection of unknown security vulnerabilities (0Day). Moreover, it is more automated and has better performance in finding new security vulnerabilities.