Geolocated Data Generation and Protection Using Generative Adversarial Networks

Hugo Alatrista-Salas, Peter Montalvo-Garcia, Miguel Nunez-del-Prado, Julián Salas

Producción científica: Capítulo del libro/informe/acta de congresoContribución a la conferenciarevisión exhaustiva

1 Cita (Scopus)

Resumen

Data mining techniques allow us to discover patterns in large datasets. Nonetheless, data may contain sensitive information. This is especially true when data is georeferenced. Thus, an adversary could learn about individual whereabouts, points of interest, political affiliation, and even sexual habits. At the same time, human mobility is a rich source of information to analyze traffic jams, health care accessibility, food desserts, and even pandemics dynamics. Therefore, to enhance privacy, we study the use of Deep Learning techniques such as Generative Adversarial Network (GAN) and GAN with Differential Privacy (DP-GAN) to generate synthetic data with formal privacy guarantees. Our experiments demonstrate that we can generate synthetic data to maintain individuals’ privacy and data quality depending on privacy parameters. Accordingly, based on the privacy settings, we generated data differing a few meters and a few kilometers from the original trajectories. After generating fine-grain mobility trajectories at the GPS level through an adversarial neural networks approach and using GAN to sanitize the original trajectories together with differential privacy, we analyze the privacy provided from the perspective of anonymization literature. We show that such ϵ -differentially private data may still have a risk of re-identification.

Idioma originalInglés
Título de la publicación alojadaModeling Decisions for Artificial Intelligence - 19th International Conference, MDAI 2022, Proceedings
EditoresVicenç Torra, Yasuo Narukawa
EditorialSpringer Science and Business Media Deutschland GmbH
Páginas80-91
Número de páginas12
ISBN (versión impresa)9783031134470
DOI
EstadoPublicada - 2022
Evento19th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2022 - Sant Cugat, Espana
Duración: 30 ago. 20222 set. 2022

Serie de la publicación

NombreLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volumen13408 LNAI
ISSN (versión impresa)0302-9743
ISSN (versión digital)1611-3349

Conferencia

Conferencia19th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2022
País/TerritorioEspana
CiudadSant Cugat
Período30/08/222/09/22

Huella

Profundice en los temas de investigación de 'Geolocated Data Generation and Protection Using Generative Adversarial Networks'. En conjunto forman una huella única.

Citar esto