Open AccessAn ensemble feature selection approach using hybrid kernel based SVM for network intrusion detection system
Author(s) -
Gaddam Venu Gopal,
G. Rama Mohan Babu
Publication year - 2021
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v23.i1.pp558-565
Subject(s) - support vector machine , intrusion detection system , feature selection , computer science , artificial intelligence , kernel (algebra) , pattern recognition (psychology) , feature (linguistics) , polynomial kernel , kernel method , data mining , machine learning , mathematics , linguistics , philosophy , combinatorics
Feature selection is a process of identifying relevant feature subset that leads to the machine learning algorithm in a well-defined manner. In this paper, anovel ensemble feature selection approach that comprises of Relief Attribute Evaluation and hybrid kernel-based support vector machine (HK-SVM) approach is proposed as a feature selection method for network intrusion detection system (NIDS). A Hybrid approach along with the combination of Gaussian and Polynomial methods is used as a kernel for support vector machine (SVM). The key issue is to select a feature subset that yields good accuracy at a minimal computational cost. The proposed approach is implemented and compared with classical SVM and simple kernel. Kyoto2006+, a bench mark intrusion detection dataset,is used for experimental evaluation and then observations are drawn.