ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

IMPLEMENTATION OF SEQUENTIAL SVM CLASSIFIER TO IMPROVE RESPONSE TIME AND DETECTION RATE

Journal: International Journal of Engineering Sciences & Research Technology (IJESRT) (Vol.4, No. 6)

Publication Date:

Authors : ; ; ;

Page : 636-643

Keywords : Classification; Intrusion Detection System(IDS); Multiclass SVM (MCSVM); Neural Network(NN); Sequential Minimal Optimization (SMO); Support Vector Machine(SVM).;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

It is important to protect the assets we exchange on network that may be harmed by malicious activities. For detection of malicious activities we use Intrusion Detection System (IDS). Focus of this paper is on IDS using SVM. SVM is used as it has property of high scalability and high speed of classification which proves it efficient for IDS. The survey result shows problem of low detection rate and high response time when using traditional SVM. The problem of surveyed literature is overcome by implementing SVM-SMO model proposed in this paper and using appropriate pre-processing method. The model optimizes the Lagrange multiplier and finds support vectors which SVM algorithm uses for classification. The SVM-SMO model in this paper is implemented for sequential as well as parallel approach. The weight updating module is used by Sequential approach which prioritizes SVM classifier for improved performance. The experimental result shows improvement of 3.94% and 1.85s in detection rate and response time by implementing SVM-SMO model in sequential approach as well as improvement of 8.91s in response time using parallel approach.

Last modified: 2015-07-12 02:02:00