A Differentially Private Trajectory Publishing Mechanism Based on Stay Points

Trajectory data contains abundant of spatiotemporal information, publishing unprotected trajectories may disclose individual privacy. Recently, researchers have proposed differential privacy to protect users’ privacy when publishing trajectory. However, existing works tend to introduce additional noise when add Laplacian noise. To solve this problem, we propose a differentially private trajectory mechanism publishing based on stay points. Firstly, TF-IDF is used to estimate the importance of each stay point and applied to exponential mechanism as a utility function. Additionally, important stay points can be selected by exponential mechanism and assigned corresponding privacy budget based on the value of TF-IDF. Furthermore, noise which added to each protected stay point, is generated from two-dimensional Laplacian via sampling distance and angle between adjacent points. Experiments on two real trajectory data sets show that our proposed mechanism has high data availability while satisfying the privacy protection level.