multitask-vulberability-detection

This tool, VulSATD, employs multi-task learning to enhance the detection of vulnerabilities and self-admitted technical debt (SATD) in code, leveraging the relationships between these two types of code quality issues. Utilizing a deep learning approach based on CodeBERT, it facilitates experimentation with different training modes, including single and multi-task setups, and allows researchers to evaluate the impact of weighted loss on detection performance. Despite its ambitious design, the evaluation indicates that sharing information between SATD and vulnerabilities did not yield a significant performance improvement in automatic detection.