TRAJECTORY DIFFERENTIAL PRIVACY PROTECTION MECHANISM BASED ON SEMANTIC LOCATION CLUSTERING

Aiming at the problems of being vulnerable to semantic attacks and having low data availability in the current trajectory data privacy protection schemes, a trajectory differential privacy protection scheme based on semantic location clustering is proposed. Firstly, the semantic distances between various positioning points in the trajectory are estimated by sorting out the logical relationships of different semantic concepts. Then, the clustering algorithm is used to generate clustering results with members having high semantic similarity for the trajectory data set as anonymous sets. Secondly, the differential privacy exponential mechanism is utilized to select representative positions with a lower possibility of privacy leakage from the clustering results to anonymize the sensitive points in the original trajectory, which achieves good privacy protection effects while avoiding large information losses.