DYNAMIC PACKET FILTERING USING MACHINE LEARNING METHODS

Abstract

With the emergence of the Internet, cyber-attacks and threats have become significant issues. Traditional manual network monitoring and rule-based packet filtering methods have become labor-intensive and less effective in combating attacks. Filtering packets based solely on payload and pattern matching is also inefficient. There is a need for a dynamic model capable of learning packet filtering rules. This article proposes a packet filtering model using Neural Networks. After developing the model classified with training and validation data, it can be utilized to support dynamic packet filtering. The proposed model allows filtering packets not only based on static rules but also considering IP packet attributes and rules learned by the model in advance. The model takes into account payloads and other IP packet attributes for filtering. It can automatically update firewall rules to enhance security.