A support vector machine approach to the development of an intrusion detection system

Abstract

This paper demonstrated the use of support vector machine (SVM) model to develop an Instruction Detection System (IDE) which detects attacks by classification in wireless local area networks. The implementation was done on real time environment where network packets captured were subjected to SVM model for classification as either an attack or a normal data. The classifications were performed by separating the data into different clusters using a hyperplane. Waterfall model software development  methodology was used to develop the intrusion detection system application and  implementation was carried out with java programming language. The model predictive ability was evaluated by modelling an attack type and comparing the results with a  standard benchmark, 99.78% and 98.89% were obtained for the detection of normal and attacks when tested with 5500 packets. The results suggest an improved detection efficiency and false alarm rate.

Key words: Support Vector Machine, Classification algorithm, Network Security, Network Packets