Balancing trajectory privacy and data utility using a personalized anonymization model

With the widespread use of location-based services (LBS), the number of trajectories gathered by location service providers is dynamically growing. On the one hand, mining and analyzing these spatiotemporal trajectories can help to work out a mobile-related strategic planning; on the other hand, knowledge of each trajectory can be used by adversaries to identify the user's sensitive information and lead to an unpredictable harm. The concept of trajectory k-anonymity extends from location k-anonymity that has been widely used to address this issue. The main challenge of trajectory k-anonymity is the selection of a trajectory k-anonymity set. However, existing anonymity methods ignore the trajectory similarity and direction, assuming that it has little impact on privacy. Thus, it cannot provide a preferable trajectory k-anonymity set. In this paper, we propose to use trajectory angle to evaluate trajectory similarity and direction, and construct an anonymity region on the basis of trajectory distance. Considering the various preference settings on the proportion of trajectory privacy and data utility in different scenarios, we propose a personalized anonymization model to select the trajectory k-anonymity set. Experiment results prove that our method can provide an effective trajectory k-anonymity set under various proportions of trajectory privacy and data utility requirements, while the efficiency just reduces a little.