Software Vulnerability Root Cause Analysis Method Based on Data Analysis

With the continuous advancement of information technology, software systems have grown significantly in scale, leading to increasingly prominent security challenges. Fuzzing, a widely adopted software security testing technique, can automatically generate inputs to trigger abnormal program behaviors, thereby uncovering potential vulnerabilities. Although fuzzing is capable of producing inputs that cause program crashes-known as Proof of Concept (PoC)-it does not directly reveal the underlying root causes of these crashes. In practice, security analysts are often faced with a large number of crash samples and must manually debug each one to trace the root cause of the vulnerability and determine the corresponding fix location. This process is typically time-consuming and labor-intensive. Automating root cause analysis (RCA) for vulnerabilities has therefore become a critical and urgent challenge. Existing RCA approaches still suffer from significant limitations in terms of practicality: they tend to be time-intensive and often produce results with limited accuracy, making them insufficient for guiding real-world vulnerability remediation. As a result, inefficient vulnerability analysis and prolonged patch cycles remain common issues. To address these challenges, this paper proposes a novel automated root cause analysis method, named DARCA. DARCA transforms the problem of vulnerability root cause analysis into data analysis. By statistically examining and correlating the runtime behavior of key program variables, it identifies those variables that are highly associated with program crashes-i.e., the root cause variables. DARCA can rapidly and accurately identify the critical factors responsible for crashes without the need for extensive manual debugging, thereby significantly improving the efficiency of vulnerability analysis. We implemented a prototype of DARCA and evaluated it on 20 known historical vulnerabilities drawn from 10 open-source software projects. Experimental results demonstrate that, compared to state-of-the-art approaches, DARCA achieves an average improvement of $\mathbf{7 6 \%}$ in vulnerability localization accuracy and an 81% increase in analysis speed, showcasing its substantial advantages in practical root cause analysis tasks.