Research On Privacy Protection Mechanism Of Privacy Labeled Trajectory Data Publishing

With the explosive growth of location-aware devices and wireless communication,it is easy to capture the moving trajectories of users and beneficial to both users and service providers to release trajectory data.More and more applications are mining these data to enrich people's lives such as geo-social network,location-based marketing,and mobile health.Published data for data mining is beneficial to both users and the research community.However,the publishing trajectory data may pose privacy threats to users even after easy anonymization.In such data-driven applications,a trajectory is usually composed of one label and a sequence of spatio-temporal points.In this paper,we deeply researched trajectory data publishing,and divided the privacy protection mechanisms into partition-based and differential privacy-based privacy protection mechanism and proposed corresponding privacy protection methods for both types of mechanism.For the partition-based privacy protection mechanism,this paper designed the(l,?,?)-privacy model to defend against three kinds of attacks: record link,attribute link and similarity attack.We designed a novel perturbation method that selects an addition or subtraction operation based on a defined utility function,thereby avoiding utility loss caused by excessive suppression in traditional methods.For the differential privacy-based privacy protection mechanism,we used differential privacy for the first time to implement location and sensitive label privacy protection in trajectory data publishing.For the first time,we noticed the utility value of outliers in the trajectory,which was preserved and protected by differential privacy.We designed a novel graph-based differential privacy algorithm,which includes clustering and generalization of trajectories,graph-based differential privacy on points,and publishing labeled trajectories.In this paper,for the two privacy protection schemes proposed,our performance studies based on a comprehensive set of real-world data and privacy analysis show that two privacy protection schemes can effectively protect data privacy while protecting data utility.