A three-way trajectory privacy-preserving model based on multi-feature fusion

Abstract

The prevalent method in trajectory privacy protection through publishing k-1 similar trajectories alongside a target trajectory, often relies on a single feature, which can compromise the balance between privacy, data utility, and processing efficiency. Addressing this, our study introduces a nuanced three-way decision model that integrates multiple trajectory features: staying areas, average velocity, and distance. This model begins by filtering candidate trajectories through staying areas and average velocity. Subsequently, it employs a three-way decision-making process based on a novel dynamic trajectory warping (DTrW) distance metric, which obviates the need for trajectory pre-alignment and mitigates information loss. This process classifies trajectories into accepted, rejected, or pending categories, with the final set compiled through a combined metric of average speed and staying areas. Comparative experiments on a classic real trajectory dataset validate the model’s superiority, demonstrating enhanced privacy protection, improved data utility, and increased efficiency over single-feature-based approaches and other k-anonymity models.