Reinforcement Learning for Dynamic Web Firewall Policy Optimization
Zaynab Sai'd, Sherifdeen Kayode · 2025
The dynamic and ever-evolving nature of cyber threats necessitates adaptive and intelligent security mechanisms for web firewalls. This research investigates the application of Reinforcement Learning (RL) algorithms to optimize web firewall policies in real-time, enabling proactive defense against sophisticated attack patterns. By continuously learning from network traffic and security events, the RL agent adaptively tunes firewall rules to minimize false positives and false negatives, improving both security efficacy and system performance. The study explores various RL techniques, such as Q-learning and deep reinforcement learning, to evaluate their effectiveness in dynamically adjusting firewall configurations under diverse and unpredictable cyber-attack scenarios. Experimental results demonstrate that RL-driven firewall policy optimization significantly enhances the responsiveness and robustness of web security infrastructure compared to static rule sets, paving the way for more autonomous and resilient cyber defense systems.