Research On Hybrid Trust Model And Key Techniques For VANETs

The emergence of Vehicular Ad hoc Networks(VANETs)facilitates the further improvement of road safety and traffic efficiency.Before vehicles can coordinate with each other and have these positive effects on the traffic,the issue of trust management should be addressed as the trust is a prerequisite for the coordination.That is,in an open system such as VANETs,where most of the vehicles are unknown to one another,there is a need to provide means to build up trust.Fitting trust to VANETs is intrinsically complex,and the following issues need to be addressed.First,how to define the trust metric according to the chosen trust factor.Since one of the typical applications in VANETs is vehicle platooning,it is practically important to find what kind of target driving behavior helps the formulation of vehicle platooning and measure the deviation between the target behavior and the actual behavior.Second,how to calculate the trust in vehicles through the entity-centric trust management model.The essential characteristics of VANETs are random topology,self-organization and dynamic connections;therefore,it is hard to establish reliable and stable relationship among vehicles.Third,how to quickly and correctly compute the trust in reports on a traffic event through the data-centric trust management model.The computed results may be outdated since the traffic event itself is dynamic.Moreover,vehicles usually have difficulty in gathering enough evidence to address the credibility of reports due to the essential characteristics of VANETs,resulting in inaccurate results.Fourth,how to defend against outside and inside attacks.Outside attackers attempt to eavesdrop on the communication channel,tamper with messages,or even crash the system,and inside attackers try to manipulate the trust values of targeted vehicles.The main contributions of this dissertation are listed as follows.(1)The target driving behavior is defined by Linked Vehicle Model(LVM),which sets a benchmark for driving behavior evaluation.LVM is an idealized car-following model,and it features collective motion whereby each vehicle applies behavior rules of cohesion and separation to keep a desired gap with the vehicle ahead.Compared with other realistic or idealized car-following models,it achieves comprehensively optimal microscopic and macroscopic performance.Simulated results show that it achieves maximum mean speed and maximum traffic flow in the virtual road network.The results also show that it usually performs well in terms of pollutant emission and fuel consumption.Then the driving behavior trust metric(DBTM)scheme based on LVM is presented,and it measures the deviation of vehicles from the benchmark.The application of DBTM in a real dataset attains reasonable distribution of trust values.(2)An entity-centric trust model named Implicit Web of Trust in VANETs(IWOTV)is proposed to evaluate the trustworthiness of vehicles.The “implicit” comes from the fact that the web is derived from dynamic and opportunistic interactions among vehicles,and is independent from the explicit fast-changing topology of VANETs.The problem of the trust evaluation is transformed into the problem of obtaining the stationary probability vector via link analysis.IWOT-V is implemented via two algorithms: Bayes Trust and Vehicle Rank.The former computes the local trust.It regards the local trust as a random variable,which denotes the unobservable trustworthiness of vehicles.According to techniques from Bayesian statistics,the random variable is continually updated based on accumulated evidence or observations.The latter computes the global trust.It can be explained with the credit-welfare model that a successor vehicle earns credit from the predecessor and receives welfare from the system for having been trustworthy.The simulation results show that IWOT-V can accurately identify trusted and untrusted nodes if enough local trust information is collected.(3)A computing framework for Internet of Vehicles(Io V)named Quick Fake News Detection(Qc FND)is proposed to detect the fake news quickly and correctly.Qc FND exploits the technologies from Software-Defined Networking(SDN),edge computing,blockchain,and Bayesian networks.Qc FND consists of two tiers: edge and vehicles.The edge is composed of Software-Defined Road Side Units(SDRSUs),which are extended from traditional Road Side Units(RSUs)and host virtual machines such as SDN controllers and blockchain servers.The SDN controllers help to implement the load balancing on Io V.The blockchain servers accommodate the reports submitted by vehicles and calculate the probability of the presence of a traffic event,providing time-sensitive services to the passing vehicles.With the accumulation of traffic reports,the Bayesian network is exploited to continually update the posterior probability of traffic events via establishing a causal network representing the probabilistic relationships among hypothesis variables and information variables.Extensive simulations and experiments show that Qc FND achieves the better performance compared with other solutions.(4)Distributed Voting Scheme(DVote)for the trust management is designed to secure the system from outside and inside attacks.To defend against outside attacks,the permissioned blockchain is used as the bulletin board,where only certified vehicles can post ballots according to the observations of serving vehicles.To defend against inside attacks,the votes are encrypted with the Paillier cryptosystem,the homomorphic property of which enables DVote to compute the sums of votes without revealing the content of individual votes.Moreover,the game theory is exploited to prove that rational vehicles wouldn’t like to be inside attackers and the manipulation of votes is unlikely to happen.Therefore,the proposed system doesn’t depend on the assumption that most vehicles are honest.Hyperledger Fabric is chosen as the experimental platform and DVote is implemented on it.All in all,the proposed hybrid trust model consists of the four modules mentioned above,i.e.DBTM,IWOT-V,Qc FND and DVote,fulfilling the calculation of the trustworthiness of entities and data.The research of this dissertation has established theoretical and technological base for the improvement of trust management in VANETs,and it can be applied to the deployment of VANETs for the improvement in road safety and traffic efficiency.