Evaluation of Feature Selection Techniques in Intrusion Detection Systems Using Machine Learning Models in Wireless Ad Hoc Networks

To better understand existing intrusion detection systems (IDSs) and the techniques adopted to develop an effective system, this chapter presents an exhaustive literature review. It elucidates the framework used in the present research work to detect wormhole attacks in WANs. The chapter also discusses the simulation environment used to build the IDS. It analyzes the relationships between the features in the network layer, and thereby presents the influence of feature selection in the IDS. Two machine learning models are used for feature selection, namely, random forest method and PCA method. The chapter demonstrates the model developed for intrusion detection using one‐class SVM technique. It is concluded that the IDSs designed using PCA + K‐means cluster classifier IDS is very effective in the detection of wormhole attacks, whereas PCA + one‐class SVM IDS is good in the detection of wormhole attacks in WANs.