PFcrowd: Privacy-Preserving and Federated Crowdsourcing Framework by Using Blockchain.

Chen Zhang,Yu Guo,Hongwei Du,Xiaohua Jia

DOI: https://doi.org/10.1109/iwqos49365.2020.9212891

2020-01-01

Abstract:Crowdsourcing is a promising computing paradigm that utilizes collective intelligence to solve complex tasks. While it is valuable, traditional crowdsourcing systems lock computation resources inside each individual system where tasks cannot reach numerous potential workers among the other systems. Therefore, there is a great need to build a federated platform for different crowdsourcing systems to share resources. However, the security issue lies in the center of constructing the federated crowdsourcing platform. Although many studies are focusing on privacy-preserving crowdsourcing, existing solutions require a trusted third party to perform the key management, which is not applicable in our federated platform. The reason is that it is difficult for a third party to be trusted by various systems. In this paper, we present a secure crowdsourcing framework as our initial effort toward this direction, which bridges together the recent advancements of blockchain and cryptographic techniques. Our proposed design, named PFcrowd, allows different crowdsourcing systems to perform encrypted task-worker matching over the blockchain platform without involving any thirdparty authority. The core idea is to utilize the blockchain to assist the federated crowdsourcing by moving the task recommendation algorithm to the trusted smart contract. To avoid third-party involvement, we first leverage the re-writable deterministic hashing (RDH) technique to convert the problem of federated task-worker matching into the secure query authorization. We then devise a secure scheme based on RDH and searchable encryption (SE) to support privacy-preserving task-worker matching via the smart contract. We formally analyze the security of our proposed scheme and implement the system prototype on Ethereum. Extensive evaluations of real-world datasets demonstrate the efficiency of our design.