Hybrid Supervised Machine Learning-based Intrusion Detection System of Internet of Things

Abstract

The Internet of things (IoT) is an embedded network of networked computing devices found in everyday objects, allowing them to send and receive data and makes our lives comfortable. Therefore, IoT is a very important technology and at the same time security has become a challenge, due to several issues, including poor interoperability, security flaws, privacy concerns, and lack of industry standards. Cyber-attacks on the IoT may impact security and privacy. To address the IoT-related security issue, nowadays, intrusion detection system (IDS) and machine learning technique are commonly used. In addition, to speed up the detection process, optimal features selection techniques are incorporated into the IDS. This paper proposes an efficient IoT IDS based on a hybrid feature selection and supervised machine learning-based technique. To examine the robustness of the proposed IDS, experiments are carried out on two datasets, i.e. NSL-KDD and UNSW-NB15. The proposed IDS achieves a detection accuracy rate of 99.3% on the NSL-KDD dataset, and 99.4% on UNSW-NB15 dataset. The proposed IDS was also compared to the deep learning-based IDS and found out that the proposed IDS achieves better performance in term of accuracy and learning runtime.