A Privacy-Preserving and Identity-Based Personalized Recommendation Scheme for Encrypted Tasks in Crowdsourcing

Abstract

Personalized task recommendation can guarantee that tasks are pushed to the right workers, and thus, requesters gain better-quality output from a crowdsourcing system. Requesters' tasks and workers' interests exposed to the remote crowdsourcing platform in the form of plaintext have raised serious privacy concerns. Encrypting tasks and interests is a feasible solution to protect the privacy of both the requesters and the workers. However, data encryption renders the existing crowdsourcing task recommendation techniques ineffective. To address this challenge, in this study, we transform the personalized task recommendation problem into a task access control and keyword-based search problem for encrypted tasks. On the basis of this idea, we first develop a new technique called multi-authority attribute-based searchable encryption by equipping the searchable capacity for Lewko and Waters's multi-authority attribute -based encryption. Then, we utilize the new technique to construct a secure and personalized recommendation scheme in crowdsourcing, which achieves accurate personalized task recommendation in a privacy-preserving manner by a seamless combination of attribute-based encryption and searchable encryption. We provide rigorous security proof and thorough security analysis for the proposed scheme. Extensive experiments demonstrate the correctness and practicality of the proposed scheme.