An Efficient Firewall Packet Filtration Using Ensembled Neural Networks Model

Abstract

Firewalls are one of the most important network security tools for protecting networks from external threats. According to packet attributes and security regulations, firewalls filter and act on the packets. Because attacks are becoming more complicated, and more attacks are being made all the time, it is very difficult to make rules manually. Machine-learning algorithms decrease human effort and develop more sophisticated and efficient rule sets. In this study, we developed an ensemble artificial neural network model for web packet filtering. The proposed method is based on the Internet Firewall Dataset. The model under consideration was trained on ten distinct subsets of the dataset and achieved a remarkable accuracy rate of 99.8%. This represents a 3.4% improvement in accuracy over conventional neural networks trained on the entire dataset, which has 96.4% accuracy