Differentially private publication of streaming trajectory data

Abstract

User-generated trajectories (e.g. during traveling) can be leveraged to offer value-added services (e.g. smart city policy formulation), but there are also privacy implications. For example, information about the routes or destinations obtained from such published trajectories can be used to profile and identify users, including during contact tracing in pandemics (e.g., COVID-19) or the monitoring of demonstrations (e.g., surveillance). However, existing trajectory publishing algorithms generally rely on batch processing platforms, and rarely pay attention to real-time privacy protection processing in streaming scenarios. Therefore, we propose a stream processing framework containing two modules for spatio-temporal data. This framework is designed to achieve high data utility, while effectively ensuring the preservation of privacy in the published results. The first module is TSP, which concurrently receives real-time queries from individuals and releases new sanitizing trajectories. The second module is VCR comprising three algorithms based on differential privacy to facilitate the publication of the distribution of position statistics. Our experiments on real-world datasets demonstrate that our framework can effectively guarantee privacy with high data utility, when the appropriate parameter configuration is chosen. In addition, compared with the baseline algorithm Ht-publication, our group-based algorithm AGn-publication achieves better data accuracy in terms of visitor counts at the same level of privacy protection.