Title: A novel mutual feature CRF for intrusion detection system

Authors: Jinny S. Vinila; J. Jayakumari

Addresses: Department of Computer Science and Engineering, Department of Electronics and Communication Engineering, Noorul Islam University, Kumaracoil, Kanyakumari, Tamil Nadu, India ' Department of Computer Science and Engineering, Department of Electronics and Communication Engineering, Noorul Islam University, Kumaracoil, Kanyakumari, Tamil Nadu, India

Abstract: Intrusion detection system (IDS) is a system that monitors the network to find out the suspicious activity there by stopping the disruption that can be caused by intruders. Disruption caused by intruders can be stopped by having effective IDS. Intrusion detection system has three works, i.e., it continuously monitors network, compares with knowledge base and gives alarm to the administrator. The effectiveness of the system resides on the knowledge base. Preparing knowledge base is the most required part. This can be simply thought as a data analysis system. Data mining algorithms are applied to develop the knowledge base. Mining algorithm alone will not produce best knowledge base and so an additional step is required to sharp the algorithm. In this paper, we propose a mutual information feature selection algorithm with conditional random field, which produces high performance, less false positive IDS.

Keywords: conditional random field; feature selection; data mining; intrusion detection systems; IDS; information gain; mutual feature CRF; network security.

DOI: 10.1504/IJENM.2015.073873

International Journal of Enterprise Network Management, 2015 Vol.6 No.4, pp.299 - 311

Received: 21 Nov 2014
Accepted: 08 Jan 2015

Published online: 27 Dec 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article