Application of Machine Learning in the Optimization of Firewall Parameters Based on Attack Detection
Dragan Jeftić, Petar Čisar, Kristijan V. Kuk, Vladica S. Stojanović · IPSI Transactions on Internet Research · 2025
This paper explores the optimization of firewall parameters for attack detection using machine learning techniques, focusing on improving network security in dynamic environments. Traditional firewall systems often face limitations in detecting malicious traffic due to static rule sets and high false positive rates, particularly in real-world scenarios with evolving attack patterns. To overcome these challenges, this study applies neural networks with the rectified linear unit activation function (ReLU), which enables precise attack detection and real-time firewall policy adjustments. The proposed 5-5-4 neural network model, tested using real-world datasets, achieved an accuracy of 96.3%, outperforming alternative configurations. The analysis evaluated three scenarios: normal conditions, active attacks, and post-policy adjustment, confirming the effectiveness in enhancing detection and mitigation capabilities. The results highlight the potential of machine learning, particularly neural networks, as a robust tool to improve network security. This approach enables future integration of IoT and real-time threat monitoring.