Trust Assessment Model Based on a Hierarchical Decision-Making Process in IoT Sensor Networks

The Internet of Things (IoT) and its security issues have been gaining increasing attention in recent years due to the increasing number of threats and attacks against IoT devices. Specifically, false data injection attacks aim at tampering with the sensed data that reflect the normal system operation state. Therefore, a time-efficient and accurate approach to assessing sensor trustworthiness is necessary. To this end, we propose T-LHDM (a trust model based on low and high decision-making processes), a sensor state trust model based on a Markov chain to represent sensor state transitions in normal operating conditions. To demonstrate the model, we first observed sensed data in a hierarchical decision-making process by intentionally exploiting the sensor data’s temporal and spatial correlations. Then, the consistency between the precomputed set of sensor state transitions and the observed ones was used to detect corruptions and assess the sensor trustworthiness. Our experiments using two real-world datasets demonstrated that the proposed model accurately assessed the sensor trustworthiness with an average accuracy of 95%. We also stress the significance of considering the sensor’s spatiotemporal correlations to assess the trust of the sensed data, which has been neglected in previous studies. To this end, we conducted a comparative evaluation against a contemporary temporal correlation-based method and three well-known correlation tests. The results showed that T-LHDM outperformed all the competing methods.