ZebraLancer: Crowdsource Knowledge atop Open Blockchain, Privately and Anonymously

We design and implement the first private and anonymous decentralized crowdsourcing system ZebraLancer. It realizes the fair exchange (i.e. security against malicious workers and dishonest requesters) without using any third-party arbiter. More importantly, it overcomes two fundamental challenges of decentralization, i.e. data leakage and identity breach. First, our outsource-then-prove methodology resolves the critical tension between blockchain transparency and data confidentiality without sacrificing the fairness of exchange. ZebraLancer ensures: a requester will not pay more than what data deserve, according to a policy announced when her task is published through the blockchain; each worker indeed gets a payment based on the policy, if submits data to the blockchain; the above properties are realized not only without a central arbiter, but also without leaking the data to blockchain network. Furthermore, the blockchain transparency might allow one to infer private information of workers/requesters through their participation history. ZebraLancer solves the problem by allowing anonymous participations without surrendering user accountability. Specifically, workers cannot misuse anonymity to submit multiple times to reap rewards, and an anonymous requester cannot maliciously submit colluded answers to herself to repudiate payments. The idea behind is a subtle linkability: if one authenticates twice in a task, everybody can tell, or else staying anonymous. To realize such delicate linkability, we put forth a novel cryptographic notion, the common-prefix-linkable anonymous authentication. Finally, we implement our protocol for a common image annotation task and deploy it in a test net of Ethereum. The experiment results show the applicability of our protocol and highlight subtleties of tailoring the protocol to be compatible with the existing real-world open blockchain.