Software Vulnerability Prediction

Software vulnerability prediction aims to automatically identify weaknesses in software code before exploitation, improving software security. Current research focuses on leveraging deep learning models, particularly large language models (LLMs) and graph neural networks (GNNs), often incorporating techniques like multi-task learning and self-instruction to enhance accuracy and generalization across different programming languages and vulnerability types. These advancements are crucial for mitigating the ever-increasing threat of software vulnerabilities, enabling more efficient and effective security testing and ultimately improving software reliability and safety.