Privacy-Preserving Worker Recruitment Under Variety Requirement in Spatial Crowdsourcing

Abstract

With the rapid growth of spatial crowdsourcing applications, more and more people are benefiting from it. The idea of spatial crowdsourcing is recruiting a set of workers to finish the spatial tasks. Existing worker recruitment mechanisms do not consider the variety requirement, which is easy to meet if the Spatial Crowdsourcing (SC) platform has full knowledge of the data of each worker. Since the SC platform is not fully trusted, workers are concerned about the privacy of their data. To prevent information leaks, workers’ data needs to be specially processed before it can be sent to untrusted platforms for task assignment. The data specially processed by existing privacy-preserving processing methods cannot be used directly to complete such variety tasks with high quality. To solve this problem, we propose a new variety optimization method based on the classical local differential privacy (LDP) mechanism. It can efficiently select the sets of workers with variety of categorical attributes while providing privacy protection for workers. In addition, we also propose a two-step LDP perturbation protocol that can improve the optimization result in the case of uneven distribution of worker attributes. Extensive experiments on synthetic and real datasets show that our methods can efficiently select variety worker subset with better task quality than baseline and close to optimal selection results.