Hidden Markov Trust for Attenuation of Selfish and Malicious Nodes in the IoT Network

The exposure of IoT nodes to the internet makes them vulnerable to malicious attacks and failures. These failures affect the survivability, integrity, and connectivity of the network. Thus, the detection and elimination of attacks in a timely manner become an important factor to maintain network connectivity. Trust-based techniques are used in understanding the behavior of nodes in the network. The proposed conventional trust models are power-hungry and demand large storage space. Succeeding this Hidden Markov Models have also been developed to calculate trust but the survivability of the network achieved from them is low. To improve survivability, selfish and malicious nodes present in the network are required to be treated separately. Hence, in this paper, an improved Hidden Markov Trust (HMT) model is developed, which accurately detects the selfish and malicious nodes that illegally intercept the network. The proposed model comprises the Learning Module which aims to understand the behavior of nodes and compute trust using HMT with the expected output. The probability parameters of the HMT model are derived from the data flow rate and the residual energy of the nodes. Next, in Decision-Module, the actual nature of the node is obtained with the help of the evaluated node’s likelihood functions. If the node is selfish and is close to crashed state then, is isolated from the routing function, while the selfish node with sufficient energy is immediately destroyed from the network. On the other hand, malicious nodes are provided with a time-based opportunity to reset themselves before being knocked down. Finally, if the node is legitimate, then the function continues smoothly. At last, the Path-Formation-Module establishes the trusted optimal routing path. Further, comparative analysis for attacks such as black-hole, grey-hole, and sink-hole has been done and performance parameters have been extended to survivability-rate, power consumption, delay, and false-alarm-rate, for different network sizes and vulnerability. Simulation result on average provides a 10% higher PDR, 29% lower overhead, and 17% higher detection rate when compared to a Futuristic Cooperation Evaluation Model, Futuristic Trust Coefficient-based Semi-Markov Prediction Model, Opportunistic Data Forwarding Mechanism, and Priority-based Trust Efficient Routing using Ant Colony Optimization trust models presented in the literature.