Automated Vulnerability Detection in Source Code Using Deep Representation Learning

Researchers developed a convolutional neural network model that can automatically detect vulnerabilities in C source code using deep learning techniques. The model was trained on datasets from Draper Labs and NIST, achieving higher recall than previous work while maintaining high precision and demonstrating effectiveness on real Linux kernel vulnerabilities.