Finding granular features using rough-PSO in IDS

Abstract

Most of the existing IDS use all the features in network traffic to evaluate and look for known intrusive pallerns. Unfortunately, such system suffers a lengthy detection procedure. Serious implication may incur to a host computer or network due to delay in diagnosis. Feature reduction improves the speed of data manipulation and classification rate by reducing the influence of noise. Besides, selecting important features from input data leads to a simplification of a problem, faster and more accurate detection rates. The purpose of this paper is to investigate the effectiveness of the Rough Set and Particle Swarm (PSG) in feature selection. Support Vector Machine (SVM) was used as a classifier. Data used in this experiment was originally obtained from dataset created by DARPA in the framework ofthe 1998 Intrusion Detection Evaluation Program. Six significantfeatures were proposed by Rough-PSG.