Abstract:
This project delves into the critical need for automated source code security checks in today's modern software development process. The software development process in the industry is growing more complex day by day, the challenge of finding security weaknesses inside source code is also becoming hard. My work, "Vulnerability Detection in Source Code Using AI," offers an AI-driven method to detect and guide users for potential security problems in application source code early in the development phase. The main objective is to build an intelligent system that learns from deep learning and then uses its learning to analyse application source code and pinpoint where is the vulnerabilities. I am focusing on issues like vulnerable code injection points, insecure object handling by the code, improperly handling user data, and the other misuse of programming functions. The system learns by studying labeled dataset and from a teacher model of both secure and vulnerable code, allowing it to recognize patterns linked to insecure coding inside of an application. The overall process of my application includes preparing different code samples, extracting key features and labeling, training deep learning models from specialized teacher model, and assessing the system's accuracy and performance. The end goal of my project after successful completion is, to lessen the need for manual application source code review for application developers, educate the users with security best practice and boost the overall security inside application development process. While the current system successfully identifies several common vulnerabilities, there's room for improvement. Future work includes expanding language support and integrating the system into realtime development tools.
