Supervised Machine Learning Based Network Intrusion Detection System for Internet of Things

Abstract

The Internet of Things (IoT) is an innovative invention that can combine physical object to the Internet with an ability to transfer and access of the data through Internet, however with the rapid growth in the application and services of the IoT, the scope of network attack is also increasing exponentially. To secure data, device and IoT network, there is a need of an efficient, secure and accurate Intrusion Detection System (IDS). IDS basically monitors network and system activities and raises alarm when anything deviated from its normal behaviour is found. Classical intrusion detection system follows rule based detection approaches that fail to detect zero day or unknown attack is not suitable for dynamic and insecure IoT environment. This paper mainly proposes an efficient method with uniform detection system based on supervised machine learning technique by using Random Forest classifier. Also two different datasets, NSL-KDD and KDDCUP99 with minimal feature sets have been used that give lightweight attack detection strategy for IoT network. Simulation of proposed method with theses datasets has 99.9 percentage accuracy in intrusion detection with less amount of time and energy.