Preserving Privacy With Homomorphic Encryption For Reward-Based Spatial Crowdsourcing

Spatial crowdsourcing has become a popular paradigm for performing complicated spatiotemporal tasks due to the ubiquity of mobile devices.However,workers who participate in a crowdsourcing project are generally required to disclose their locations,which may lead to serious privacy threats.Unfortunately,providing a rigid privacy guarantee is incompatible with ensuring a high task acceptance rate in most existing crowdsourcing solutions.Hence,this paper proposes a crowdsourcing framework based on homomorphic encryption.The key idea is to tune the reward for performing each task to the workers' preferences to attain a high acceptance rate.The first step in the framework is to interrogate the workers' preferences using a cryptographic protocol that fully preserves the location privacy of the workers.Based on those preferences,two different approaches to reward assignments have been proposed to ensure the rewards are distributed optimally.A theoretical analysis of the privacy protection inherent in the framework demonstrates that the proposed framework guarantees the worker's location privacy from adversaries including the requester and crowdsourcing server.Further,experiments based on real-world datasets show that the proposed strategies outperform existing solutions in terms of task acceptance rates.This paper proposes a novel framework,SecRSC,for reward-based crowdsourcing.A cryptographic protocol based on homomorphic encryption is designed to privately investigate worker preferences for performing specific crowdsourcing tasks.This paper uses pairwise relatively prime numbers as a tag for each task to take advantage of the unique products and factors prime numbers provide to fully preserve the privacy of worker locations.It is believed the proposed cryptographic protocol could also be applied to other scenarios,such as private elections or private questionnaires.Two reward assignment approaches,IRR and OCF,are thus presented for adaptively specifying rewards for each task.Given a fixed budget,the ‘hot' tasks are given a relatively lower reward,while the ‘outliers' that might not be accepted are given attractive returns.Experimental results on real-world datasets show that,by adaptively tuning the reward for each of the tasks,the proposed reward assignment approaches significantly outperform the uniform assignment approach which is used in worker selected mode to determine the TAR.