Research On Security Mechanisms And Key Technologies In Vehicular Networks

In recent years,the ever-increasing number of vehicles pose serious challenges to existing transportation systems,e.g.,frequent traffic congestions and traffic accidents,which seriously threaten people's lives and property.Vehicular networks are regarded as promising solutions for the above problems.By establishing wireless communications channels among vehicles and network infrastructures,vehicles can effectively obtain the real-time traffic conditions and make driving decisions in advance,thereby improving the traffic safety and efficiency.However,as a complex network containing multiple elements,such as people,vehicles,base stations,etc.,there may be various security risks in it,jeopardizing the normal operation of the network and even the traffic safety and efficiency.This dissertation focuses on the security issues in the application of vehicular networks.Through taking full use of the various kinds of data generated in vehicular networks(e.g.,vehicle sensing data,communications data,and user data),specific security threats or malicious attacks can be prevented.In this article,the data-driven security architecture in vehicular networks is proposed at first.Then,after analyzing the security threats in typical vehicular applications,e.g.,online car-hailing,autonomous driving,and car sharing,the corresponding countermeasures are proposed and validated.The main contributions of this dissertation include:1.Data-driven security architecture for vehicular networks applicationsBig data have played important roles in wireless network optimization and user experience improvement.As another important indicator,the network security can also been improved by use of big data.In this part,main data types existing in vehicular networks are summarized at first,followed by their potential value in maintaining cybersecurity.Then,the data-driven security architecture is proposed,which includes basic functions of various network elements(e.g.,vehicles,base stations)and their interaction processes.Furthermore,by analyzing requirements of this architecture,including the security,privacy and performance,design goals and challenges are specified.Finally,typical use cases are briefly introduced to illustrate the feasibility of this architecture.This part provides basic ideas and methods for analyzing and solving specific security issues.2.Mobility-based Sybil detection algorithm in vehicular networksThis part focuses on the Sybil attack,which may exist in a common application,i.e.,the online car-hailing.To solve this problem,a mobility-based Sybil detection scheme is proposed.Firstly,Sybil attackers in vehicular networks are mainly divided into three levels according to their capabilities,including general attackers,attackers with location forgery,and attackers with collusion,whose behaviors are modeled respectively.Then,based on their behavioral features,corresponding countermeasures are designed and validated:(1)By collecting and analyzing the vehicle mobility trajectories,the similarity feature vector of two trajectories is calculated.Then,machine learning classification algorithms are used to detect two vehicles whose mobility behaviors are too close.Therefore,the general attackers are removed;(2)Base stations are used to issue location certificates for vehicles,based on which the subjective logic theory is utilized to evaluate the credibility of locations uploaded by vehicles.In this way,attackers with location forgery can be effectively detected;(3)According to the times and time intervals that two vehicles coexist near a base station,the proximity level between them is calculated.Then,the community detection algorithm is used to cope with attackers with collusion.Finally,a real-world vehicle traj ectory dataset is used to validate the effectiveness of these algorithms.The results show that the proposed scheme can effectively mitigate the Sybil attack at all levels.3.Blockchain-based decentralized vehicle trust management schemeThis part focuses on the trust issues between vehicles.In autonomous driving scenario,vehicles can receive messages from neighboring vehicles to obtain local traffic conditions.However,due to the stranger relationship among vehicles,received messages are not fully trusted.To solve this problem,this research utilizes the Bayesian inference model,so that the receiver can comprehensively analyze the information from multiple sources to judge the credibility of each message.Based on these judgments,receivers generate ratings for messages and upload them into the nearby base stations.These base stations form a blockchain network,which is responsible for the calculation,storage,updating and maintenance of the trust value for each vehicle.In addition,a novel blockchain consensus scheme suitable for vehicular trust management is designed,which enables base stations with more ratings to generate and broadcast new blocks faster.Therefore,the data security,consistency,and timeliness are ensured.Simulation results show that the proposed system can effectively help vehicles to evaluate the credibility of received messages and,in the meanwhile,meet the key performance requirements of the vehicular networks.4.Smart contract-based vehicle access control schemeThis part focuses on the vehicle access control problem in the car sharing application scenario.Traditional centralized access control systems are faced with some serious threatens,e.g.,the single point of failure and lack of mutual trust.Therefore,the blockchain network and the smart contracts running on it are applied in this research,aiming at establishing a decentralized and trusted vehicle access control system.Thanks to the smart contract,key access control procedures,e.g.,the authentication and authorization,can be operated safely and transparently in decentralized nodes.Compared with centralized servers,the blockchain network uses multiple nodes for data storage and program execution,which can greatly reduce the possibility of the system being intruded by attackers.Therefore,blockchain and smart contract provide a secure and trusted platform for the interactions among vehicles and individuals.Furthermore,the Reinforcement Learning algorithm is utilized to optimize the order selection strategy for vehicles.Using this strategy,veh icles can flexibly choose whether to accept an order request or not according to the system states,e.g.,the expected earnings and remaining time,so as to maximize the long-term gain of the vehicle owner.Finally,the designed system and proposed algorithm are validated on a vehicular network simulation platform through two steps.Firstly,the delay of the access control system is evaluated to verify its feasibility in car sharing applications.Secondly,the Q-learning method is used to optimize order selection strategy.The simulation results show that the proposed method outperforms the traditional Greedy algorithm.