An Integration of PSO-based Feature Selection and Random Forest for Anomaly Detection in IoT Network

Abstract

The most challenging research topic in the field of intrusion detection system (IDS) is anomaly detection. It is able to repeal any peculiar activities in the network by contrasting them with normal patterns. This paper proposes an efficient random forest (RF) model with particle swarm optimization (PSO)-based feature selection for IDS. The performance model is evaluated on a well-known benchmarking dataset, i.e. NSL-KDD in terms of accuracy, precision, recall, and false alarm rate(FAR) metrics. Furthermore, we evaluate the significance differencesbetween the proposed model and other classifiers, i.e. rotation forest (RoF)and deep neural network (DNN) using statistical significance test. Basedon the statistical tests, the proposed model significantly outperforms otherclassifiers involved in the experiment.