Traj-MergeGAN: A Trajectory Privacy Preservation Model Based on Generative Adversarial Network

Nowadays, with the rapid development of location-based services, individual trajectory data is collected for various traffic related applications. However, while we are benefiting from these services, the trajectory data may contain lots of private information and privacy issues need to be carefully handled. In this paper, we propose a deep learning model named Traj-MergeGAN, which can generate synthetic trajectory from original trajectory. The generated trajectory can not only protect individual privacy, but also maintain data quality for other downstream applications. Furthermore, we conduct overall experiments to prove the advantages of our model.