Design And Simulation Of A Trusted Edge Computing System For Internet Of Vehicles

IoV(Internet of Vehicles)is an open integrated network system consisting of vehicles,wireless networks and roadside units.It plays a vital role in the intelligent transportation system.IoV can provide vehicles with road conditions,collision warnings and so on.IoV also has the advantages of reducing traffic accidents and easing traffic congestion.However,IoV’s working requires vehicles to have high computing capability.Edge computing is accomplished by offloading computing tasks to the edge of a network.This can reduce the time delay of completing computing tasks and alleviate the pressure of computing and bandwidth on the center servers.On the one hand,vehicles in a certain range can form an edge network to complete the computing tasks together in the IoV,which can reduce the demand for the computing capacity of a single vehicle.On the other hand,because the neighbor vehicles can sense more road data,offloading the computing task to the neighbor vehicles also can reduce the overhead of data transmission and delay of task completion.However,due to the openness and dynamics of the IoV,IoV edge computing is conducted in an untrusted environment:edge computing tasks may be offloaded to an untrusted vehicle,which may result in the task not being completed in time.In addition,when a task executing vehicle returns results to the task assigning vehicle,the results may be obtained by malicious vehicles.Thus,the malicious vehicles can obtain the private information of the task executing vehicle.Currently,most of the existing researches on IoV edge computing focus on edge node allocation and task division.Few efforts have been made on the trust and privacy of vehicles.Therefore,the thesis focuses on three aspects in the context of IoV edge computing:how to establish a trusted IoV edge computing system,trust management mechanism and data privacy protection mechanism in the IoV edge computing system.The main contents include:(1)Designing a trusted IoV edge computing system.By adding functional modules such as trust evaluation,task record storage,vehicular data privacy protection,and task offloading in the vehicle nodes and RSU in the IoV,the system can provide functions like identity authentication,trust evaluation,task record storage,task offloading,vehicular data privacy protection etc.,and also can prevent networking attacks such as collusion praise attack,bad-mouthing attack,on-off attack and inference attack.The simulation results show that the designed trusted edge computing system can meet the functional requirements of task offloading,trust evaluation vehicular data privacy protection etc.,and is also scalable and reliable.(2)Solving the problem of selecting trusted vehicles for computing and offloading in the designed system.The thesis proposes a trust management mechanism for vehicles based on blockchain.The mechanism consists of trust evaluation and task record storage.The task record storage uses blockchain as a distributed storage platform to store task records.The trust evaluation uses an improved Bayesian inference to compute direct trust between vehicles,box-plot and DBSCAN to remove malicious evaluation,prevent collusion praise and bad-mouthing attack,and compute the indirect trust between vehicles.Moreover,it uses reinforcement learning to compute the comprehensive trust value of vehicles according to the number of direct interactions and indirect interactions.The simulation results show that the proposed trust management mechanism in the designed system can accurately identify malicious vehicle nodes,and prevent collusion praise attack,bad-mouthing attack and on-off attack.(3)Solving the problem of privacy leakage when the task executing vehicles return results in the designed system.Currently most of the IoV edge tasks are machine learning tasks(such as vehicle image recognition,road flow prediction,etc.).The thesis proposes an adaptive data privacy protection mechanism based on differential privacy for machine learning tasks.By adaptively cutting gradient,setting the learning rate and assigning weights according to the ratio of noise and gradient when the machine learning models are aggregated,the proposed mechanism can effectively protect the privacy of the task executing vehicles who executes the machine learning tasks.The simulation results show that this mechanism can improve the accuracy and convergence speed of the machine learning model while protecting the data privacy of the task executing vehicles.