Privacy-preserving and verifiable online crowdsourcing with worker updates

Abstract As a novel problem-solving paradigm, crowdsourcing has been emerged aiming at performing truthful data aggregation and addressing problems that are hard for one organization. However, in reality, these answers submitted by distributed workers can be regarded as their intellectual properties because of professional skills and intensive computation overhead. Besides, it may contain highly sensitive information and should not be directly released without protection. Besides, the worker skill estimation is only considered into plaintext scenario but no longer suitable to ciphertext setting. To address these challenges, we propose a novel privacy-preserving and verifiable online crowdsourcing protocol (PVOC) for workers in the same group while preserving their answers privacy. Besides, PVOC also supports worker dynamic adding and revocation for different classification tasks with a minimum computation overhead. Finally, we introduce a verification mechanism to identify and update the worker skills of participants. Thus, since no decryption operations are involved in PVOC, our design is efficient and lightweight. Security analysis demonstrates that PVOC can guarantee the workers’ privacy without accuracy loss. Furthermore, we evaluate and show its effectiveness and practicability on three real datasets MNIST, CIFAR-10 and CIFAR-100.