LLM-based Vulnerability Detection

The emerging Large language models (LLMs) are increasingly used as a revolutionised tools in many industry. In cyber security, the LLMs can offer innovative ways to detect, identify, and mitigate vulnerability through their natural language processing capability. This work aims to fine tune pre-trained LLMs to detect anomalies and vulnerability by analysing vast amounts of data. A ChatGPT 3.5 model based vulnerability detector were developed that can conduct common vulnerability analysis. Experimental results demonstrate the effectiveness of proposed solution.