Title: A study of feature reduction techniques and classification for network intrusion detection

Authors: Meenal Jain; Gagandeep Kaur

Addresses: Department of CSE&IT, JIIT, Noida Sector 62, Noida – 201309, India ' Department of CSE&IT, JIIT, Noida Sector 62, Noida – 201309, India

Abstract: The size of network data increasing tremendously day by day. This huge amount of data contains large number of attributes which need to be analysed for particular application. In this paper, three feature reduction techniques such as, principal component analysis (PCA), artificial neural network (ANN), and nonlinear principal component analysis (NLPCA) have been used to analyse the significance of such attributes. Three newly reduced datasets from the original benchmark dataset Coburg intrusion detection dataset (CIDDS, 2017), have been created. To evaluate the performance of such reduced datasets in terms of information_gain (IG) with actual data, two metrics namely, sensitivity_mismatch_measure (SMM) and specificity_mismatch_measure (SPMM) have been used. Results show that ANN and NLPCA reduces the original dataset to approximately 29% and 57%, respectively while maintaining high information gain. Also as compared to PCA, both methods outperform in eliminating attributes to retain the essential information of the dataset.

Keywords: PCA; principal component analysis; ANN; artificial neural network; NLPCA; nonlinear principal component analysis; decision tree; K-nearest neighbour; SVM; support vector machine; Naïve Bayes; SMM; sensitivity_mismatch_measure; SPMM; specificity_mismatch_measure; Information_gain.

DOI: 10.1504/IJAIP.2024.142664

International Journal of Advanced Intelligence Paradigms, 2024 Vol.29 No.2/3, pp.133 - 149

Received: 09 Mar 2019
Accepted: 11 Jun 2019
Published online: 15 Nov 2024 *

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