Hybrid Learning Approach of Combining Cluster-Based Partitioning and Hidden Markov Model for IoT Intrusion Detection

Abstract

Internet of Things (IoT) is a global network that connects various types of objects "things" via internet. It becomes a core technology for various applications and more and more embedded within our daily lives and businesses. As the technology grows and evolves a number of issues will arise and be focused on in IoT, Security is one of the central issues in IoT in the last decade. However, most of today's IoT intrusion detection systems suffer from high false alarms rate with moderate accuracy and detection rates when it's not able to detect all types of IoT intrusions correctly. To overcome this problem, hybrid techniques are used. In this paper, hybrid learning approach combining partitioning clustering techniques with Hidden Markov Model (HMM) is proposed. Experimental results show that the proposed approach using K-Medoids has improved the detection rate as well as decreased the false positive rate.