Trajectory Data Publication Based on Differential Privacy

Analyzing trajectory data can provide people with a higher quality of life. However, publishing directly will leak privacy. The authors propose publication method based on differential privacy (TDDP). TDDP consists two stages. In the location generalization stage, firstly, locations at each timestamp are clustered into classes by k-means++ algorithm, and then representative class is selected using exponential mechanism. generalized design sampling mechanism to form trajectories. sampled from under different timestamps avoid generation non-semantic ensure that trajectories resist filtering attacks. experimental results show released achieve good balance between protection availability.