DEIT: Dempster Shafer Theory‐based edge‐centric Internet of Things‐specific trust model

Abstract

The potential benefits of edge-centric Internet of Things (IoT) come with increased security threats and because the edge devices may be compromised into sending or forwarding wrong information. To negate the effect of wrong information on inference, such messages need to be identified. Identification and dropping of messages in the network as early as possible would minimize the communication overhead. Trust values of nodes help in the early identification of such messages. In this work, the Dempster Shafer Theory (DST)-based edge-centric IoT-specific Trust Model (DEIT) is designed to compute trust to include the specific characteristics of IoT. Edge servers use DEIT to compute trust value of edge devices in the network. In DEIT, DST combines the available trust messages. DIET also uses the degradation variable, credibility variable, punishment variable, re-compensating variable, and pardoning variable to compute final trust value. The computed trust value assists the edges in taking correct decisions. Simulation results demonstrate that it can immediately capture two major types of misbehavior, namely, information drop and modification under both sparse and dense traffic conditions. Through scenarios in the results section, the depiction of various possible misbehaviors has been discussed. Scenario 1 and 2 discuss a situation where a malicious edge device misbehaves in two ways, namely, dropping and modification of information, respectively.In scenario 3 and 4, situations are discussed when the trust opinion of the trusted and distrusted devices are neutralized with time using the degradation factor to illustrate and compare changes in trust values in two different situation.

Evaluating Trust Management Frameworks for Wireless Sensor Networks

“…Bhargava et al [29] proposed the "Dempster Shafer Theory (DST)-based edge-centric IoT-specific trust model (DEIT)" for WSNs. The DEIT mechanism enables edge servers to compute the trustworthiness of edge devices based on their behavioral metrics such as message drop rate and message modification rate.…”

Trust Evaluation and Decision Based on D-S Evidence Theory: Early Models and Future Perspectives

“…The demand for computational skills also increases. As the number of elements in the recognition framework increases [34], the calculation volume when using combination rule increases exponentially. Limited by the lack of sensor node computing, storage, and energy supply capacity [54], it is difficult to run algorithms with high computational complexity for a long time.…”