AI-Driven Code Vulnerability Detection System Using Deep Learning Techniques

This research paper proposes an AI-driven code vulnerability detection system using Deep Learning and Natural Language Processing techniques for intelligent source code analysis. The proposed framework integrates preprocessing, tokenization, feature extraction, Convolutional Neural Networks (CNN), and Transformer-based architectures to automatically identify vulnerabilities such as SQL Injection, Cross-Site Scripting (XSS), Buffer Overflow, insecure API usage, and weak authentication mechanisms. Benchmark datasets including Devign, VulDeePecker, and Juliet Test Suite are utilized for model training and evaluation. Experimental analysis demonstrates improved detection accuracy, lower false positive rates, and better contextual understanding compared to traditional rule-based scanners. The proposed framework contributes toward intelligent cybersecurity automation, secure software engineering, and AI-assisted software auditing.