Multi-class SVM based network intrusion detection with attribute selection using infinite feature selection technique

Abstract

An intrusion detection mechanism is a software program or a device that monitors the network and provides information about any suspicious activity. This paper proposes a multi-class support vector machine (SVM) based network intrusion detection using an infinite feature selection technique for identifying suspicious activity. Single and multiple classifiers generally have high complexity. To overcome all the limitations of single and multiple classifiers, we used a multi-class classifier using an infinite feature selection technique, which performed well with multiple classes and gave better results than other classifiers in terms of accuracy, precision, recall, and f_score. Infinite feature selection is a graph-based filtering approach that analyses subsets of features as routes in a graph. We used a standard dataset, namely the UNSW_NB15 data set generated by the IXIA perfect-storm tool in the Australian Centre for Cyber Security. This dataset has a total of nine types of attacks and 49 features. The comparative analysis of the manuscript work is done against eight different techniques, namely, hybrid intrusion detection system (HIDS), C5, one-class support vector machine, and others. The proposed work gave better simulation results using the 2015a Matlab simulator.

Subject Classification:

• 15A06
• 15A09
• 18A10
• 17B10
• 17B15
• 18C10
• 18C05
• 18D35

Keywords:

• Anomaly detection
• Support vector machine
• Infinity feature selection
• Multiclass support vector machine
• Intrusion detection system
• Machine learning

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.