Research On Threat Detection And Location Privacy Preservation Technologies In The Edge-end Environment Of The Internet Of Vehicles

With the Internet of Vehicles(Io V),vehicles can build connections with other vehicles,pedestrians,infrastructure,and even all the connected things.Io V can leverage highly connected vehicles and Location-Based Service(e.g.,fine-grained spatial query)to build a large number of intelligent transportation system applications including autonomous driving and intelligent vehicle infrastructure cooperation,etc.Connected vehicles and fine-grained Location-Based Service(LBS)not only bring the advanced intelligent transportation system but also location privacy leakage and potential threats.By analyzing the location privacy of vehicles or attacking vehicles,malicious users can percept and attack the key infrastructure in the physical world,which endangers national security.The Io V is highly decentralized,thus the threat detection and location privacy preservation mechanisms in Io V should also be built in a distributed manner.Nowadays,since the computation resources at the edge nodes(edge)and vehicles(end)of Io V dramatically increase,edge-end collaboration can be utilized to construct distributed threat detection and location privacy preservation services in Io V.However,leveraging edge-end collaboration to build threat detection and location privacy protection mechanisms for Io V still faces three challenges.First,geographical dynamic threats in Io V scenarios are hard to detect.The moving of vehicles indicates they face threats characterized by geographic dynamics.However,traditional static threat detection mechanisms are hard to detect threats that will change along with the location.Second,the intrusion detection system is hard to be built for a specific area.Because the driver has his/her own residence and workplace,the vehicle usually moves in some fixed areas.To train intrusion detection models for a specific area,traditional intrusion detection systems require vehicles to upload sensitive intrusion samples to the server.Traditional intrusion detection system lacks privacy protection and incentive mechanisms,thus vehicles are not willing to participate in the building of intrusion detection system.Third,the location privacy of the fine-grained spatial query is hard to protect.A large number of vehicles frequently send shape-unconstrained search areas to the LBS provider to judge whether its Point of Interest(e.g.,charging pile)is located in the target area.The LBS provider can obtain the sensitive location information of the vehicle during the querying process.However,traditional high-overhead privacy-preserving coarse-grained spatial query mechanisms cannot support search areas with highly complex shapes,thus they cannot protect the privacy of the fine-grained spatial query process.This thesis carried out innovative researches to address these challenges.The main work and contributions are summarized as follows:1.Micro-blockchain and edge-end collaboration based geographic dynamic cyber threat intelligence sharing in Io V.A micro-blockchain structure is firstly proposed for geographic dynamic Cyber Threat Intelligence(CTI)sharing in Io V.The micro-blockchain structure consists of one macro-blockchain and multiple micro-blockchains.Each micro-blockchain is run by edge nodes and deployed in a specific area.When the vehicle moves from one place to another,the micro-blockchains can be nested to form a larger micro-blockchain to generate CTI for vehicles in a larger area.Second,the timely generation of CTI is the key to detect potential threats,thus we proposed a federated learning and non-cooperative game based CTI generation mechanism.The proposed CTI generation mechanism enables edge nodes to encourage vehicles to provide more computation resources for CTI generation.Experimental results show that the proposed CTI sharing system has a good performance in detecting geographic dynamic threats.Microblockchain structure with three micro-blockchains is 8.67 times faster than the traditional single blockchain structure in terms of auditing CTI blocks.The proposed CTI generation mechanism is 4.12 times faster than the traditional solution in terms of generating CTI.2.Federated learning and game theory based intrusion detection for the Io V.To address the challenge of building intrusion detection systems for a specific area,a federated learning based distributed intrusion detection framework is firstly proposed.In the newly proposed framework,by organizing vehicles to run federated learning,highly distributed edge intrusion detection servers can train intrusion detection models for vehicles in a privacy-preserving manner.Second,reputation models for edge intrusion detection servers and vehicles are proposed based on subject logic theory.With the newly constructed reputation models,the federated learning process is modeled as a multi-leader multi-follower game,which can encourage edge intrusion detection servers and vehicles with higher reputation values to participate in building the intrusion detection system.The simulation results show that the edge intrusion detection server with a higher reputation value can obtain more intrusion samples from the vehicle with a higher reputation value,which can increase the detection rate of the model in such server.3.Image-based privacy-preserving fine-grained spatial query for edge-enabled Io V.In order to leverage edge nodes to provide privacy-preserving fine-grained spatial query service for the Io V,an image-centric spatial query framework is firstly proposed to generate fine-grained results for the highly complex search area.Second,a content-searchable image encryption algorithm is proposed to build image-centric privacy-preserving fine-grained spatial query.Security analysis and performance evaluations show that the proposed fine-grained privacy-preserving spatial query can resist various attack methods and generate search results in 60 ms,which is much faster than traditional mechanisms.To sum up,this thesis proposed threat detection and location privacy protection mechanisms for Io V in the edge-end collaboration environment.The experiments carried out based on the public dataset confirmed the security,effectiveness,and superiority of the proposed mechanisms.This thesis provides theoretical support and important reference for building the secure Io V.