Towards Standardized Mobility Reports with User-Level Privacy

TL;DR

This paper proposes a standardized, privacy-preserving framework for mobility reports using differential privacy, implemented as open-source software, and evaluates its effectiveness on real datasets.

Contribution

It introduces a novel standardized mobility report with differential privacy guarantees and provides an open-source implementation for practical use.

Findings

Limiting user contributions minimally affects geospatial distributions.

Reducing user contributions significantly decreases noise-induced errors.

The approach balances privacy and data utility effectively.

Abstract

The importance of human mobility analyses is growing in both research and practice, especially as applications for urban planning and mobility rely on them. Aggregate statistics and visualizations play an essential role as building blocks of data explorations and summary reports, the latter being increasingly released to third parties such as municipal administrations or in the context of citizen participation. However, such explorations already pose a threat to privacy as they reveal potentially sensitive location information, and thus should not be shared without further privacy measures. There is a substantial gap between state-of-the-art research on privacy methods and their utilization in practice. We thus conceptualize a standardized mobility report with differential privacy guarantees and implement it as open-source software to enable a privacy-preserving exploration of key aspects of mobility data in an easily accessible way. Moreover, we evaluate the benefits of limiting user contributions using three data sets relevant to research and practice. Our results show that even a strong limit on user contribution alters the original geospatial distribution only within a comparatively small range, while significantly reducing the error introduced by adding noise to achieve privacy guarantees.