Ant colony optimization and feature selection for intrusion detection

Abstract

Network intrusion detection gained a lot of attention from the security expert. Intrusion detection system has been designed for the purpose detecting attack and comprises of detection method that can be anomaly based or it can be signature based. These detection method, however, highly depends on the quality of the input features. Supervised learning approach for the detection method finds the relationship between the feature and its class. Therefore, irrelevant, redundant, and noisy features must be eliminated before applying supervised algorithm. This can be done by feature selection method. In this paper ant colony optimization has been applied for feature selection on KDD99 dataset. The reduced dataset is validated using support vector machine. Results show that accuracy of the SVM is significantly improved with reduced feature set.