TDSF: Trajectory-preserving method of dual-strategy fusion with differential privacy in LBS

Abstract

When the public utilizes location-based services (LBS), a large amount of trajectory data is generated, and their location information is constantly exposed. However, providing trajectories to LBS without additional protection may result in the leakage of location privacy and correlation privacy in the trajectory. Most current methods only protect the location privacy of trajectories by adjusting the allocation of privacy budgets, without combining multiple strategies to protect location privacy and correlation privacy. These methods also struggle to balance data availability and privacy for trajectories. To address the above challenges, we propose a trajectory-preserving method of dual-strategy fusion with differential privacy (TDSF). Specifically, one strategy is used to protect the correlation privacy between sensitive locations, and the other is used to protect the non-sensitive locations. We use the trained transfer correlation matrix to extract sensitive locations in a trajectory that require correlation protection. The remaining locations introduce less noise as they involve minimal privacy disclosure, thus maintaining data availability. Finally, we also designed a privacy budget allocation strategy that is suitable for this dual-strategy fusion scenario. Strict security analysis shows that the mechanism we propose can well protect the location and correlation privacy of the trajectory. The experimental results on real data sets further demonstrate the advantages of this mechanism in data availability and confidentiality.