Research On Trust Model Of Internet Of Things Based On Recommendation Filtering Algorithm

With the development and improvement of technologies such as 5G,the Internet of Things(IoT)is entering a stage of rapid development,bringing many conveniences to people.However,the substantial expansion of IoT also brings some security issues.In order to improve the security of IoT,we can use trust models to evaluate the trustwor-thiness of nodes in IoT,and then isolate nodes with low trustworthiness to reduce the impact of malicious attacks within the network.Besides,the combination of trust mod-els and traditional security mechanisms such as intrusion detection and access control can also further promote the secure development of IoT.There are still some problems in the research on trust models of IoT.On the one hand,most of the nodes in IoT have limited resources and heterogeneous functions,so trust models in traditional e-commerce and peer-to-peer networking(P2P)fields cannot be directly applied to IoT.On the other hand,there may be malicious nodes hijacked by attackers in these nodes.The launching of trust related attacks by malicious nodes will inevitably reduce the accuracy of trust evaluation.At present,many trust models have not taken effective measures to avoid the negative impact of malicious nodes on trust evaluation.In response to the above problems,this dissertation studies a trust model that applies to IoT and has high resistance toward trust related attacks.The specific work is summarized as follows:(1)We propose an architecture for the trust model which is based on the trust third party(TTP),and combine with a specific trust evaluation process to elaborate on the respective roles of nodes and TTP in trust evaluation.The introduction of TTP provides a safe and reliable trust evaluation environment and reduces the energy consumption of nodes.(2)We propose an adaptive trust model based on the recommendation filtering al-gorithm and divide the calculation of a node's trust value into three parts:direct trust evaluation,recommendation trust evaluation and synthesis trust evaluation.In the di-rect trust evaluation,the time decay function,sliding window mechanism and penalty factor are introduced to the direct trust evaluation method based on Bayesian inference so as to speed up the convergence speed of trust evaluation.In the recommendation trust evaluation,we propose a recommendation filtering algorithm that combines direct trust,similarity and confidence to reduce the negative impact of trust related attacks on the recommendation trust evaluation.In the synthesis trust evaluation,we propose an adaptive weight calculation method that combines direct trust and recommendation trust to strengthen the adaptivity of the trust model.(3)In response to the above work,we perform a large number of experiments based on the ns-3 simulator and verify that the architecture based on TTP can significantly reduce energy consumption.At the same time,we verify that the proposed trust model can quickly evaluate the trust value of nodes,and the proposed trust model has resistance toward attacks,which can effectively reduce the negative impact of malicious nodes on trust evaluation.