DCONST: Detection of Multiple-Mix-Attack Malicious Nodes Using Consensus-Based Trust in IoT Networks

The Internet of Things (IoT) is growing rapidly, which allows many smart devices to connect and cooperate with each other. While for the sake of distributed architecture, an IoT environment is known to be vulnerable to insider attacks. In this work, we focus on this challenge and consider an advanced insider threat, called multiple-mix attack, which typically combines three sub-attacks: tamper attack, drop attack and replay attack. For protection, we develop a Distributed Consensus based Trust Model (DCONST), which can build the nodes' reputation by sharing particular information, called cognition. In particular, DCONST can detect malicious nodes by using the K-Means clustering, without disturbing the normal operations of a network. In the evaluation, as compared with some similar models, DCONST can overall provide a better detection rate by increasing around 10% to 40%.