Feature selection for intrusion detection system in a cluster-based heterogeneous wireless sensor network

Abstract

Wireless sensor network (WSN) has become one of the most promising networking solutions with exciting new applications for the near future. Notwithstanding the resource constrain of WSNs, it has continued to enjoy widespread deployment. Security in WSN, however, remains an ongoing research trend as the deployed sensor nodes (SNs) are susceptible to various security challenges due to its architecture, hostile deployment environment and insecure routing protocols. In this work, we propose a feature selection method by combining three filter methods; Gain ratio, Chi-squared and ReliefF (triple-filter) in a cluster-based heterogeneous WSN prior to classification. This will increase the classification accuracy and reduce system complexity by extracting 14 important features from the 41 original features in the dataset. An intrusion detection benchmark dataset, NSL-KDD, is used for performance evaluation by considering detection rate, accuracy and the false alarm rate. Results obtained show that our proposed method can effectively reduce the number of features with a high classification accuracy and detection rate in comparison with other filter methods. In addition, this proposed feature selection method tends to reduce the total energy consumed by SNs during intrusion detection as compared with other filter selection methods, thereby extending the network lifetime and functionality for a reasonable period.