A Study on Adversarial Attacks for Benchmarking Deep Learning-Based Web Application Firewalls

Web Application Firewalls (WAFs) play a crucial role in protecting web applications from malicious attacks, but their effectiveness is often challenged by the evolving sophistication of adversarial training techniques. In this study, we investigate the robustness of deep learning-based WAFs by generating adversarial payloads designed to evade the detection model. These adversarial examples are utilized to train deep learning models aimed at improving the classification performance. Our results demonstrate that the integration of adversarial training significantly enhances the model's ability to identify and mitigate adversarial attack payloads, leading to a more sustainable WAF. Through comprehensive experiments, we benchmark the effectiveness of our approach, providing insights into the potential of adversarial attacks as a critical tool for evaluating and strengthening deep learning-based WAF. This work contributes to the growing field of adversarial machine learning and highlights the importance of robust evaluation frameworks in the development of secure web application infrastructures.