TCAC-GAN: Synthetic Trajectory Generation Model Using Auxiliary Classifier Generative Adversarial Networks for Improved Protection of Trajectory Data

Abstract

Mobile social networking (MSN) is gaining significant popularity owing to location-based services (LBS) and personalized services. This direct location sharing increases the risk of infringing the user’s location privacy. In order to protect the location privacy of users, many studies on generating synthetic trajectory data using generative adversarial networks (GANs) are being conducted. However, GAN generates limited synthesis trajectory data due to mode collapse problem. In this paper, we propose a trajectory category auxiliary classifier-GAN (TCAC-GAN) that generates synthetic trajectory data with improved utility and anonymity by reducing mode collapse using ACGAN. In experiments, the performance of utility and anonymity of TCAC-GAN is compared with LSTM-TrajGAN.