DL-Based XSS Attack Detection Approach Using LSTM Neural Network with Word Embeddings

Web applications are exposed to a variety of risks. Security attacks can lead to several damages including execution of malicious code, content alteration, and user session hijacking. Cross-Site Scripting (XSS) is part of these attacks allowing attackers to inject malicious scripts into vulnerable web pages. Reducing XSS risk was, and still is, one of the most ambitious targets for researchers, professionals, and industry leaders. Traditional tools are still insufficient. Deep learning techniques have improved the state-of-the-art in various fields, offering the potential to automatically learn, analyze, and predict. In this study, to improve XSS mitigation, we propose a Deep Neural Network model using LSTM architecture and word embeddings to capture temporal dependencies and contextual semantics in XSS payloads. The effectiveness of the proposed model is evaluated by analyzing its performance. The experimental results show that our model can achieve an accuracy of 99.56%