Detection of Reflected XSS Vulnerabilities Based on Paths-Attention Method

Cross-site scripting vulnerability (XSS) is one of the most frequently exploited and harmful vulnerabilities among web vulnerabilities. In recent years, many researchers have used different machine learning methods to detect network attacks, but these not achieved high accuracy recall rates cannot effectively combat XSS attacks. Designing a model that can achieve truly proactive defense against reflected has become top priority for maintaining user security at this stage. paper, we propose detection based on paths-attention method (PATS model). Firstly, converts data into an intermediate representation abstract syntax trees, then traverses tree generate multiple sets syntactic paths, them vector representations through word embedding matrices. The extracts semantic features using attention mechanisms improve training effectiveness by assigning appropriate weights paths as it learns with neural networks, which realizes transformation from passive active defense. Additionally, in dataset processing section, point out shortcomings current research datasets construct reliable composed 1000 malicious samples NIST 10,000 benign GitHub experimentation purposes. Experimental results show compared other models, rate 90.25% F1-score 81.62%, while reducing time half 30 h.