Intrusion detection using fuzzy classification and feature selection based on genetic algorithm

M. Shahrezaei,A. Sharifi,Reza Tavoli

Abstract:Nowadays, different methods are applied for the detection of abnormal attacks on computer networks. In the present study, attempts have been made to minimize errors in intrusion detection and to find its effective optimal features. Generally, an Intrusion Detection System is comprised of three parts: the first part includes sensors that collect information; the second are data processors that detect network or system status; and finally the third part presents types of responses to the attack. In this study, one part of intrusion detection system has been simulated that processes data. In the tests, use has been made of a set of intrusion detection standard data, named KDD99, serving as the evaluated data resource. Also, neural networks have been used for intrusion detection. Recommended procedure include data processing steps through analysis algorithm of main variables, selection of optimal variables and selection of optimal features through Genetic Algorithm by minimizing error of intrusion detection using Adaptive NeuroFuzzy Inference Systems.