Using neural networks in intrusion detection systems

Abstract

This paper sets out the possibility of deployment of neural networks in the most widespread open source IDS program Snort. The principle is to capture packets of network traffic and then to use neural network to increase the likelihood of identifying attack in the stream by dynamically identifying the operating systems on the individual protected hosts. Thanks to the use of neural network this system should be able to overcome some of the evasive techniques the attackers use.