A Novel Approach for Enhancing XSS Attack Detection: Optimizing Deep Neural Network with Genetic Algorithms

Deep Learning (DL) has emerged as a powerful tool for detecting and mitigating security vulnerabilities in web applications, including Cross-Site Scripting (XSS) attacks.These attacks exploit vulnerabilities in web application code, enabling malicious scripts to execute within trusted environments.This study introduces an innovative approach to enhance XSS attack detection by optimizing a Deep Neural Network (DNN) using Genetic Algorithms (GA).The GA-based optimization refines the weights and biases of the DNN, addressing challenges posed by local minima and non-convex error surfaces.Our method incorporates advanced feature extraction techniques, combining Term Frequency-Inverse Document Frequency (TF-IDF) with the N-grams method, to effectively represent contextual and lexical patterns, including obfuscated or encoded payloads.Comparative analyses with traditional Gradient Descent (GD)-based optimization techniques and alternative feature extraction methods, such as Word2Vec, highlight the significant advantages of our approach.Tested on two publicly available datasets, our approach achieved significant performance, with accuracy scores of 99.92% and 99.36%, precision of 100% and 99.91%, recall of 99.86% and 98.82%, F1-scores of 99.93% and 99.36%, ROC-AUC values of 99.99% and 99.86%, and a reduced False Positive Rate (FPR), respectively.Explainable AI (XAI) techniques, in particular SHAP (SHapley Additive exPlanations), are integrated to provide insights into the model's decision-making process, ensuring transparency and interpretability.