Impact of Centrality on Automated Vulnerability Detection Using Convolutional Neural Network

Software vulnerabilities have now become a major security threat because of the increasing size and complexity of software programs. These security flaws make vulnerability detection a prior concern. Despite the efficiency of Deep Learning (DL) based approaches, prior studies have faced challenges in maintaining detection accuracy and scalability simultaneously. In order to overcome this challenge, we propose an automated and scalable vulnerability detection technique to acquire both accuracy and scalability while searching for vulnerabilities in source code. Specifically, this method uses complex network analysis theory to effectively convert a function's source code to an image which considers program's syntactic and semantic information and then applies text convolutional neural network to detect vulnerabilities. Experimental results compared against the conventional methods report that our method achieved 5% improvement in accuracy and can detect vulnerabilities in large scale source code.