Feature reduction scheme for anomaly‐based intrusion detection in wireless networks: Building of hybrid model

Abstract

Anomaly-based intrusion detection plays a crucial and essential role in providing security to computer networks. However, some concerns still exist for the sustainability and feasibility of existing approaches over modern networks. Specifically, these concerns relate to increasing many features from a network, thus making it difficult for a user to achieve high detection accuracy. This article presents a novel and efficient feature reduction scheme for fault-based intrusion detection by efficiently handling the large feature from a network. Monarch butterfly optimization is used with a new correlation-based fitness function to reduce features. Twin support vector machine is utilized with its unique classification in the proposed scheme. The proposed method is analyzed and evaluated with NSL-KDD, CICIDS2017, AWID network security, and standard fault network datasets. The obtained results are verified using standard performance measures, that is, overall detection accuracy, F-measure, FPR, and AUC score with time complexity. The results show the superiority and stability of the proposed scheme over the existing techniques.