CATS: Conditional Adversarial Trajectory Synthesis for Privacy-Preserving Trajectory Data Publication Using Deep Learning Approaches

CATS: Conditional Adversarial Trajectory Synthesis for Privacy-Preserving Trajectory Data Publication Using Deep Learning Approaches
Jinmeng Rao, Song Gao , Sijia Zhu
GeoDS Lab, Department of Geography, University of Wisconsin-Madison, WI
International Journal of Geographical Information Science
Abstract: The prevalence of ubiquitous location-aware devices and mobile Internet enables us to collect massive individual-level trajectory dataset from users. Such trajectory big data bring new opportunities to human mobility and GeoAI research but also raise public concerns with regard to location privacy. In this work, we present the Conditional Adversarial Trajectory Synthesis (CATS), a deep-learning-based methodological framework for privacy-preserving trajectory data publication. CATS applies K-anonymity to the underlying spatiotemporal distributions of aggregated human mobility, which provides a distributional-level strong privacy guarantee. By leveraging conditional adversarial training on K-anonymized human mobility matrices, trajectory global context learning using the attention-based mechanism, and recurrent bipartite graph matching of adjacent trajectory points, CATS is able to reconstruct trajectory topology from conditionally sampled locations and generate high-quality synthetic trajectory data, which can serve as supplements or alternatives to raw data for privacy-preserving trajectory data publication. The experiment results on over 90k trajectories show that our method has a better performance in privacy preservation, characteristic preservation, and downstream utility compared with baseline methods, which brings new insights into privacy-preserving human mobility research and explores data ethics issues in GIScience.

Source code for the work entitled "CATS: Conditional Adversarial Trajectory Synthesis for Privacy-Preserving Trajectory Data Publication Using Deep Learning Approaches".

Due to the non-disclosure agreement with the data provider, we are not releasing the original individual-level GPS trajectory data but sharing the mocked individual GPS trajectory data with the same data structure and also the k-anonymized aggregated human mobility data used in the experiments.

Content Introduction

- aggregate_mobility_matrix.py: generate the aggregated mobility distribution file from individual GPS trajectories
- cats.py: provides a PyTorch implementation of CATS, including a TrajGenerator (CatGen) class and a TrajDiscriminator (CatCrt) class.
- train_cats.py: provides a training example code for CATS.
- run_cats.py: provides a inference example code for CATS.
- log_util.py: logging class.
- stmm_dataset.py: provides a Dataset class used for training the CATS.

- stmm_data/: k-anonymized aggregated mobility matrix data.
- mocked_individual_gps_data: the mocked individual GPS trajectory data samples with the same data structure with raw data