Research On Identity Recognition In Vehicle Trajectory Data

The vehicle platooning is a novel and promising technology for highway scenarios with the ability to improve road capacity and ensure travel safety.One widely acceptable form of vehicle platoon is Cooperative Adaptive Cruise Control(CACC).In this system,vehicles communicate state information with one another to maintain a coordinated,small gap between vehicles within a platoon.However,this frequently broadcasted state information to ensure the safety of the vehicle,if eavesdropped,will dramatically jeopardize the location privacy of the vehicle itself.Although in a conventional vehicle ad-hoc network,some methods based on a combination of pseudonym and mix-zone have been used to split the vehicle trajectory,by changing the vehicle pseudonym and confusing the vehicle state in the mix-zone.However,such approaches are not sufficient to ensure the location privacy of vehicles in the platoon.High-precision safety information at specific frequencies,as well as the high regularity of vehicle row behavior,introduces relatively high predictability and spatiotemporal correlation for the vehicle's trajectory.In this paper,we exploit the state information broadcast outside the mix-zone to illustrate how the attacker can infer anonymization by connecting the state information of the car.We propose an innovative deanonymization strategy using expectation maximization and maximum likelihood estimation algorithm,which can be used in various general traffic scenarios.We demonstrate the effectiveness of this deanonymization framework by extensive simulation based on both simulation-based data and real data of U.S.Highway101.Compared to traditional non-platooning transportation solutions,our strategy can achieve a 30% higher deanonymization accuracy.