Feature Reduction and Selection Based Optimization for Hybrid Intrusion Detection System Using PGO followed by SVM

Abstract

The requisition or insistence of internet (web) connectivity i.e. wireless network like WSN, MANET, cellular network, broadband increases day by day. So it is obvious that increase demand of connectivity increase the problem also i.e. safety and security. In this paper discusses the security issue or problem on connectivity network generally define as the network intrusion (malicious activity) finding system. This system has to be used for secure or protect the information data from any unwanted activities. In this paper presents the feature reduction and selection based on an optimization mechanism which followed by supervised learning classifier. This paper introduce the hybrid intrusion detection system using supervised classifier i.e. SVM followed by the optimization mechanism i.e. PGO. Every IDS system needs reduce feature data set attributes to perform efficiently and smoothly that has to be major issue for any NIDS. The hybrid optimization mechanism provide the optimal solution, plant growth optimization mechanism inspired the natural tree growth process, here make this an artificial plant growth process and apply for data set attributes and set similar condition. That optimization method provide the best fitness value for branches and leaf for an artificial plant, these branches or leaf fit for artificial plant growth or not. According to these fitness values data set attributes further classified into intruder class. In this paper present mechanism or system use NSL-KDD data set (i.e. basically intruder class attribute data sets contain DOS, PROBE, R2L and U2R intruder class) for evaluation and comparing the mechanism performance in term of accuracy and Kappa. This hybrid mechanism based on optimization decreased the false alarm rate of the system and enhance the performance.