A Layered Approach to Network Intrusion Detection Using Rule Learning Classifiers with Nature-Inspired Feature Selection

Abstract

Intrusion detection systems are meant to provide secured network computing environment by protecting against attackers. The challenge in building an intrusion detection model is to deal with unbalanced intrusion datasets, i.e., when one class is represented by a small number of examples (minority class). Most of the time it is observed that the performance of the classification techniques somehow becomes biased toward the majority class due to unequal class distribution. In this work, a layered approach has been proposed to detect network intrusions with the help of certain rule learning classifiers. Each layer is designed to detect an attack type by employing certain nature-inspired search techniques such as ant search, genetic search, and PSO. The performance of the model has been evaluated in terms of accuracy, efficiency, detection rate, and false alarm rate.