BBVD: A BERT-based Method for Vulnerability Detection

Software vulnerability detection is one of the key tasks in field software security. Detecting source code advance can effectively prevent malicious attacks. Traditional methods are often ineffective and inefficient when dealing with large amounts code. In this paper, we present BBVD approach, which treats high-level programming languages as another natural language uses BERT-based models processing domain to automate detection. Our experimental results on both SARD Big-Vul datasets demonstrate good performance proposed detecting vulnerability.