The Research Of Protecting Users' Trajectory Privacy Based On Local Differential Privacy

With the popularity of mobile smart terminal devices and the rapid development of mobile Internet,various service platforms have begun to appear in people's lives,such as Location-based Services(LBS),Mobile Crowd-Sensing,MCS),etc.While providing services to users,these service platforms also collect user's locations.Although the service platforms will give users some compensation,the leakage of users' trajectory privacy will still reduce users' participation.In order to protect the trajectory privacy of users,various technical means have appeared,such as pseudonym mechanism,K-anonymity mechanism,differential privacy,etc.Since differential privacy protect users' privacy with strong guarantee,it has been widely used in users' trajectory privacy protection.Differential privacy usually performs on statistical data on third parties or servers,so as to protect the privacy of users' trajectory.However,what if the third parties or servers are untrusted? Therefore,we consider protecting users' trajectory privacy from the users' side in this thesis,and propose mechanism of protecting users' trajectory privacy based on local differential privacy.In the process of providing services to users,some applications allow users to package and upload their trajectories offline instead of submitting location information dynamically and continuously,such as environmental monitoring.Some services require users to upload their real-time locations,such as traffic condition monitoring,route planning,and so on.For the above two situations,we have designed corresponding algorithms to protect users' trajectory privacy in this thesis.Based on this,we also consider the validity of the data submitted by the users to ensure that the third parties' or servers' utility in this thesis.For the situation that the third parties or the servers can tolerate the delay,we consider the users' trajectory as a point in high-dimensional space,and the position points on the trajectory as values on different coordinate axes in this thesis.Then,generating the total noise and assign it to different locations on the users' trajectory.Finally,users upload their trajectory in a package.For cases where users need to upload real-time locations,we consider the users' trajectory as a number of sub-trajectories of length w in this thesis,and dynamically allocate the privacy budget to each locations on the sub-trajectory,ensuring that users of any sub-trajectory with length w meets differential privacy.To improve the two algorithms,we quantify the correlation between users' trajectories,the amount of privacy leakage,and the utility of third parties or servers in this thesis.We use real world traffic trajectories of Shanghai taxis to evaluate our mechanism.The experimental results show that the algorithms provided in this thesis can provide protect users' trajectory privacy with strong guarantee,and at the same time,they ensure the validity of the data submitted by users.