SAP‐TrajWGP: A Semantic‐Aware Personalized Trajectory Privacy Protection Algorithm Based on WGAN‐GP
TL;DR
A Semantic‐Aware Personalized Trajectory Privacy Protection algorithm (SAP‐TrajWGP) based on Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN‐GP) that outperforms mainstream baseline methods in both data utility and privacy protection.
Abstract
The widespread adoption of the Internet of Vehicles (IoV) and location‐based services (LBS) has generated massive trajectory data, which drives intelligent transportation innovations but also causes severe privacy leakage risks. Existing trajectory privacy protection methods face problems of low‐quality synthetic data, insufficient semantic protection, and lack of personalization. To address these issues, this paper proposes a Semantic‐Aware Personalized Trajectory Privacy Protection algorithm (SAP‐TrajWGP) based on Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN‐GP). First, a WGAN‐GP model integrating LSTM and self‐attention mechanisms is constructed to generate high‐authenticity synthetic trajectories with spatiotemporal consistency. Second, a Bidirectional Gated Recurrent Unit (Bi‐GRU) model dynamically identifies sensitive trajectory segments, and a hierarchical differential privacy (DP) perturbation mechanism is implemented based on user‐specific privacy preferences. Experimental results on real trajectory datasets show that the proposed algorithm achieves a mean RMSE of 37.72 m, spatial and temporal JSD of 0.18 and 0.21, respectively, a TUL attack success rate of 18.4%, and an MI value of 1.86, outperforming mainstream baseline methods in both data utility and privacy protection. This research provides an effective technical pathway for secure and controllable trajectory data sharing.