Ensembling PCA-based Feature Selection with Random Tree Classifier for Intrusion Detection on IoT Network

Abstract

Technologies, applications and services of Internet of Things (IoT) are growing tremendously. This IoT blast provides an extensive choice of opportunities for consumers and manufacturer, but at the same time carriages major risks with regards to security. As more appliances and sensors become interconnected, securing them will be the major challenge. In order to make IoT objects work efficiently, hardware, software and connectivity require being secured. Less consideration on security for IoT, the connected objects may degrade the performance of services provided by the IoT network. One significant type of attack is denial of service attack (DoS) caused by manipulating handshake Transmission Control Protocol (TCP) mechanism, i.e.: TCP SYN flooding. To solve the DoS attack on IoT networks, ones use Intrusion detection system (IDS) as a potential solution. This paper proposes IDS by combining principle component analysis (PCA) feature selection technique with 3 classifier algorithms, i.e.: Random Tree (RT), K-Means, and Naïve Bayes (NB). Experimental results on IoT tesbed networks traffic dataset show that the proposed IDS using Random Tree classifier achieves the best performance in term of accuracy and energy consumption.