چكيده به لاتين
The Internet of Things (IoT) is an emerging technology which aims at facilitating human tasks by globally connecting things such as home appliances, ventilation systems, industrial and agricultural machines, sensors and vehicles. In this context, the quality of service and the trustworthiness of data being used for decision making are very important. Misbehaving users and devices may provide invalid data or evaluation of services to compromise the quality of the services being provided. Hence, a trust management system to assess trust for users or nodes, services and gathered data is deemed to be essential to every IoT ecosystem. The most common method of assessing trust in IoT applications is to estimate the trust level of the end entities (entity-centric trust) which are the nodes or the users providing feedbacks or observations. Also, in IoT applications, the data is gathered from various sensors and will be used in the actuators after the decision making process. Therefore, the trustworthiness of data (data-centric trust) is equally important and should be considered as well.
In this research, we mainly focus on assessing trust in a hybrid manner, aiming at trsut computation for both end-entities as well as data. A trust computation procedure is introduced, based on which, a computational model for assessing trust is defined that can be utilized in data collection scenarios in IoT. In our proposed model, a Bayesian learning method is used for computing the entity trust, while Dempster-Shafer Theory is exploited for data fusion and assessing data trustworthiness. As a proof of concept for our model, we conduct a case study on trust computation in a smart parking scenario, and investigate the performance of our model in the presence of misbehaving drivers and faulty parking sensors. We evaluate the convergence behavior as well as the resiliency of our method as the participating entities change their behavior. We also compare the performance of our approach with a state-of-the-art IoT trust computational model. As evidenced by the simulation experiments, our method results in a superior performance in terms of the converged values for both data trustworthiness and entity trust. It is also more resilient against misbehavior.
