Research On Model Of Creditability-based Data Fusion For The Internet Of Vehicles

With the application and extension of the technology of networked vehicles, its securityhas already become a research focus in recent years. Establishing a trust mechanism is aneffective measure to solve the security of networked vehicles, creditability model as the coreof trust mechanism, can reflect the dynamics and instability of the environment of networkedvehicles through the effective measurement and assessment information. At the same time,creditability model offers a lot more flexibility to adapt to the complex and changeableenvironment of networked vehicles.Moreover, the internet of vehicles uses data fusion technique to handle lots of data toimprove the efficiency of the system. The primary goal for the internet of vehicles is to ensurethat the sensor node for collecting traffic information can provide perceives data accurately,and timely. In order to guarantee the authenticity and reliability of the result in data fusion,this paper give a research on the problem of fabricating or tampering data during the fusionprocess of the internet of vehicles. First, this paper has deeply studied about two typicalmodels, the Beth model and the Josang model, based on analyzing the advantages anddisadvantages of the two models, combined with the internet of vehicles’ own characteristics,described and measured the relationship of direct trust and recommend trust in the internet ofvehicles, a new trusted model based on evaluating credibility of nodes is proposed. Secondly,this paper comprehensively assesses the credibility of the nodes using statistical methods andKL distance theory. Considering that trust reduces with time, it gives formula of creditabilityand construction method of the trust relationship table with the introduction of time decayfactor. According to the real-time change of trust relationship,it figures out an updatealgorithm of credibility. At last, Nodes are grouped fusion according to the credibility of themand the fusion results are evaluated. High credibility nodes are reserved while low credibilitynodes are abandoned in order to avoid their effects in data fusion.The simulation experiments approve that the model is effective, assess the credibility ofnodes accurately, identify the malicious node and behavior effectively and it proves thereliability of data fusion.