Precision Leads Recalling You! Improved Location Privacy for Shared Mobility Services
Authors: Debasree Das (University of Bamberg, Germany), Daniela Nicklas (University of Bamberg, Germany)
Volume: 2026
Issue: 4
Pages: 1041–1054
DOI: https://doi.org/10.56553/popets-2026-0158
Abstract: The rapid growth of shared micromobility, such as e-scooters and e-bikes, has transformed urban transportation, bridging the gap between public transit and first/last-mile mobility. As users and cities need information on the status and usage of such micromobility vehicles, operators publish the data using General Bikeshare Feed Specification (GBFS)-compliant APIs. These feeds are extremely useful for operational transparency and enabling third-party integration into navigation apps. However, it has raised significant privacy concerns, particularly around the fine-grained spatiotemporal data sharing, which can reveal potentially sensitive information about travel patterns, even without explicit personal identifiers. For instance, by leveraging high-precision GPS coordinates, battery levels, and timestamps, malicious actors can infer trip origins and destinations, posing a risk of membership inference attacks. Despite efforts to mitigate such risks through dynamic vehicle IDs and GBFS guidelines, the potential for privacy leakage remains. In this paper, we investigate these privacy risks in the context of micromobility data, addressing four key research questions: (1) identifying vulnerable fields in GBFS data that can leak trip trajectories; (2) validating trip origin-destination inference attacks without access to ground truth data from operators; (3) assessing the generalizability of such attacks across different cities, operators and GBFS version; and (4) proposing effective mitigation strategies. We propose a heuristic method for reconstructing trip origins and destinations using only publicly available GBFS data, without relying on vehicle identifiers or auxiliary quasi-identifiers. Our empirical analysis, conducted on data from two cities with varying sizes, shows that a significant proportion of trips can be accurately recalled, with over 80% of trip source and destination pairs identified across both cities. Furthermore, our proposed anonymization techniques, such as data generalization and removal of quasi-identifiers, can prevent up to 97% of successful attacks, ensuring privacy without sacrificing data utility.
Keywords: micromobility, location privacy, origin-destination inference attack, mitigation
Copyright in PoPETs articles are held by their authors. This article is published under a Creative Commons Attribution 4.0 license.
