VulnGPT: Enhancing Source Code Vulnerability Detection Using AutoGPT and Adaptive Supervision Strategies

In this paper, we present a novel approach to vulnerability detection in source code using a collaborative setup built on top of AutoGPT, with a controller and an evaluator AI working together. The controller oversees the evaluation process and adds a layer of self-critique to the GPT- 4 model, while the evaluator AI conducts the security assessment. By following a step-by-step interaction process, the controller prompts the evaluator AI to verify identified vulnerabilities, enabling the AI to self-correct and improve its accuracy. We discuss the results of our approach, which demonstrates the potential for effective vulnerability detection and highlights areas for improvement. Our research aims to advance the development of AI-driven security evaluation techniques to enhance the overall quality of vulnerability detection, which can be used in various areas such as IoT.