What is it about?

Location Trajectories represent a sequence of locations visited by an individual, for instance, collected through smartphones. This data is valuable for data analysis tasks, such as transport optimisation, contact tracing, or marketing. However, the trajectories can reveal sensitive knowledge such as sexual, political, or religious orientations. In this work, we show that existing approaches to protect location trajectories are not providing sufficient privacy despite claiming guaranteed privacy. Hence, this work highlights the need for more research to properly protect location trajectories.

Featured Image

Why is it important?

Our findings show that even protection approaches that rely on the de-facto privacy gold standard Differential Privacy cannot provide sufficient privacy for the individuals whose trajectories are published. Therefore, locations trajectories should not be published solely relying on such protection measures. Moreover, it is crucial that the research community develops better protection methods in order to unlock the huge potential of location trajectory analyses.

Perspectives

Many privacy-preserving publication mechanisms for location trajectories providing different types of guarantees have been proposed. In particular, there are multiple publications providing the rigorous guarantees of differential privacy, which might lead to the conclusion that the problem is solved. This work shows, however, through a significant reduction of the provided privacy level, that the existing approaches do not satisfy all privacy requirements, and therefore, highlight that more research is required. In my opinion, the main take-away of the work is that we urgently require better protection mechanisms, and we must not rely on the current state-of-the-art for trajectory protection.

Erik Buchholz
University of New South Wales

Read the Original

This page is a summary of: Reconstruction Attack on Differential Private Trajectory Protection Mechanisms, December 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3564625.3564628.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page