Time will not tell: Temporal approaches for privacy-preserving trajectory publishing

Fine-granular spatio-temporal trajectories, i.e., time-stamped sequences of locations, play a pivotal role in transport and urban analytics. However, sharing or publishing trajectory data of individuals raises concerns about location privacy given the potential for re-identification and unintentional dissemination of sensitive information. A key enabler for privacy breaches is precise temporal information. Thus, this study investigates the privacy-preserving capabilities of third-party free mechanisms protecting trajectories by exclusively targeting the temporal dimension. We compare a deterministic and a stochastic technique for shifting trajectories in time by adding an offset to each timestamp. The stochastic approach leverages a generalized version of differential privacy to render an individual's presence at any event plausibly deniable, obstructing re-identification attacks based on spatio-temporal side knowledge. Furthermore, we present a Markov chain-based speed perturbation technique that preserves dynamic patterns while obfuscating travel times and motion attributes. Using simulated re-identification attacks, we analyze privacy gains and contrast them with the utility loss. The results demonstrate a favorable utility-to-privacy ratio of the temporal techniques compared to established spatial and spatio-temporal approaches. This underlines the importance of accounting for temporal aspects in addition to spatial considerations in privacy-preserving trajectory publishing.