Dummy-Based Privacy-Preserving Mechanism Research On Trajectories

In recent years, location-based services enable us to enjoy our colorful life. Location-based services incorporate into many applications, like finding friends around me, rec-ommending restaurants around me. When enjoying the location-based services, we might reveal private information inadvertently. Once privacy attacker collects the pri-vate information, they will not use them with the best of intentions. Therefore, privacy-preserving on location-based services is a meaningful field to research.We survey the popular privacy-preserving solutions on trajectories, and target at studying the dummy-based approach. The dummy-based privacy-preserving approach is a popular technology that can protect real trajectories from exposing to attackers. Moreover, it does not need a trusted third part, while guaranteeing the quality of ser-vice. When user requests a service, dummy trajectories anonymize the real trajectory to satisfy privacy-preserving requirements.In this paper, we propose a new privacy model that includes three reasonable pri-vacy metrics:Δt-short term Disclosure, Long term Disclosure, and Trajectories Dis-tance Deviation. We also design a new algorithm named adaptive dummy trajectories generation algorithm to derive uniformly distributed dummy trajectories. Dummy tra-jectories generated by our algorithm can achieve stricter privacy-preserving require-ments based on our privacy model. The experimental results show that our proposed algorithm can use fewer dummy trajectories to satisfy the same privacy-preserving re-quirements than existing algorithms, and the distribution of dummy trajectories is more uniformly.Privacy-preserving approaches can not absolutely prevent private information re-vealing to attackers. In order to study the limitation of dummy-based privacy-preserving mechanism, we simulate privacy attackers to mine frequent patterns on mixed trajecto-ries. The experimental results show that dummy-based privacy-preserving mechanism can prevent privacy revealing to a certain extent, and the adaptive algorithm has an excellent result fending off the attacker.