Research On Trajectory Data Protection Based On Differential Privacy

With the popularity of Location-based Services(LBS),users' location information is also frequently collected,but in the trajectory data which are composed of several locations,there is a large amount of user privacy information.Once the user's trajectory is stolen by the attacker,the attacker can mine the user's privacy from the user's trajectory data,such as the address,personal preferences and other information,and then attack the user.Therefore,the trajectory data protection method has become a hot issue in the current research field of privacy protection.Most of existing trajectory data privacy protection methods adopt the strategy of adding noise to all locations,which protects the users' privacy but reduces the availability of data in varying degrees after protection.Aiming to solve this problem,we propose two methods: trajectory-differential privacy-protection method with interest region and trajectory-differential privacy-protection method with PF-Tree.In the first method,an area where a user stays long enough within a certain distance range is defined as interest region and it is mined by setting time threshold and distance threshold.And the corresponding central point of the interest region is defined as stay point,then we can generate a simplified trajectory by using stay point to represent the corresponding locations in the interest region.After that,we mine the frequent-stay points from all stay points by setting support threshold.At last,this method uses the Laplace mechanism to add noise to the frequent-stay points.After such processing,this method realizes trajectory data privacy protection under differential privacy by adding noise to a part of the locations in a trajectory.In the second method,based on the establishment of the interest regions,it mines frequent sequential patterns in stay points and uses PF-Tree to allocate privacy budget.According that,the Laplace mechanism is used to generate and add noise to the corresponding locations,and finally obtains the new trajectory data.This method makes an improvement on the first method.Instead of splitting the stay points,it fully considers the temporal sequence relationship between them.It selects the noise addition objects from the opinion of temporal sequence and reasonably allocates privacy budget.The second method improves the rationality of the noise adding method,and further improves the degree of privacy protection.Both methods in this paper are experimented on the same real data set and compared to the baseline methods.The experimental results show that the proposed methods can improve data availability under the premise of protecting the privacy of trajectory data.