Privacy preserving online matching on ridesharing platforms

Ridesharing platforms, as typical applications of spatial crowdsourcing, are becoming more and more popular in the era of mobile internet and sharing economy. One of the most fundamental issues on ridesharing platforms is to assign orders to drivers, which can be naturally modeled as online bipartite matching problem. However, conventional online matching algorithms usually lack data privacy protection mechanisms. This has become a serious issue since the spatiotemporal data of passengers is often sensitive. New policies such as EU's General Data Protection Regulation (GDPR) also enforce protection of sensitive data, which further exacerbate the privacy issues. To deal with the problems, in this paper we propose a framework based on differential privacy (DP) techniques to preserve the privacy of individuals on ridesharing platforms. Specifically, we devise a novel approach to perturb locations in online minimum bipartite matching problem and theoretically show that the performance of the perturbed matching algorithm has the same magnitude with the unperturbed one. Experiments conducted on real datasets have also shown the effectiveness of proposed framework.