Evaluation of network layer feature reduction in intrusion detection system for Wireless adhoc networks

Abstract

Wireless adhoc network is a decentralized network which has dynamic topology to newlineestablish end-to-end communication through multi-hop route operation with multiple newlinenodes cooperating themselves when source and destination nodes are not in the newlinecoverage. Due to wireless nature, it has more security issues than infrastructure newlinenetwork. Wormhole attack is one of the major security threats in a network layer. It newlinecan be launched easily even without having any prior knowledge of network. newlinePrevention mechanisms rely on only signatures. Reactive measures like intrusion newlinedetection systems detect all kinds of attacks. Most of intrusion detection systems use newlinefeatures from different layers of network. One of problem is computing a large set of newlinelayered features from ad hoc network. This research work uses three intrusion newlinedetection techniques to reduce number of wireless feature set for detection of worm newlinehole attack. They are 1. Genetic algorithm based feature reduction. 2. Principal newlineComponent Analysis (PCA) 3.Sequential Pattern Matching (SPM). newlineThe reduced feature set has been applied to principal component analysis and newlinesequential pattern matching for detection of attacks. In order to improve the detection newlinerate, fuzzy logic system has been introduced to genetic based intrusion detection newlinesystems. The performance metrics of three intrusion detection techniques have been newlineanalysed. Genetic based fuzzy methods show highest detection rate of 95.5% and newlinelowest false alarm rate of 14%. newline. newline