A Trajectory Data Publishing Scheme Integrating Semantic Feature Clustering and Differential Privacy Technology Based on the Internet of Things

Privacy protection of trajectory data is particularly important in the IoT environment. To address the problems of low data availability and susceptibility to semantic inference attacks in the current IoT-based trajectory data privacy protection scheme, a trajectory data privacy publishing scheme (SCDP) is proposed to integrate semantic feature clustering and differential privacy techniques. The trajectory data generated by IoT devices usually contains rich semantic information. First, the logical relationships between the semantic classification concepts of geolocation are sorted out to estimate the semantic distances of different location points, and then a density clustering method based on semantic distances is used to generate multiple clustering results with members having higher semantic similarity at each specified timestamp in the trajectory dataset. Second, a differential privacy exponential mechanism is used to select the alternative locations from the clustering results that are close to the center of the clusters and have a lower possibility of privacy leakage, and then privacy replacements are performed on sensitive locations in the original trajectory data after adding an appropriate amount of Laplace noise. The experimental results show that compared with other similar schemes, the proposed algorithm can effectively improve data availability and avoid large information loss while maintaining good privacy protection, which is suitable for a variety of IoT application scenarios.