A Differentially Private Framework for Spatial Crowdsourcing with Historical Data Learning

Abstract

Spatial crowdsourcing (SC) is an increasing popular category of crowdsourcing in the era of mobile Internet and sharing economy. It requires workers to be physically present at a particular location for task fulfillment. Effective protection of location privacy is essential for workers' enthusiasm and valid task assignment. However, existing SC models with differential privacy protection usually deploy only real-time location data, and their partitioning and noise additions overlaps for grids generation. Such a way may produce large perturbations to counting queries that affect success rate of task assignment and the accuracy of allocation results. In this paper, we propose a privacy framework (R-HT) for protecting location data of workers involved in SC. We use historical data learning to perform partitioning, and import real-time data directly into the second-level grid with perturbation, which realizes parallel allocation of privacy budget and a more suitable Private Spatial Decomposition (PSD) approach. Moreover, as geocast region (GR) construction, we introduce some optimization strategies, including scoring function and adaptive selection of locally maximum geocast radius. A series of experimental results on real-world datasets shows that our proposed R-HT scheme attains a stable success rate of task assignment, saves obvious performance overhead and is also fit for dynamic assignment of tasks on online SC platforms.