Personalized Data Collection Based on Local Differential Privacy in the Mobile Crowdsensing

Abstract

Mobile crowdsensing is growing in popularity by collecting environmental information from participants' mobile phones. However, the sensing data may carry sensitive information of participants so as to violate their privacy. Thus, local differential privacy (LDP) is proposed to protect participants' privacy during data collection. But most recent studies only apply LDP to the data collection without considering the participant's personal privacy preservation requirement so as to reduce the data utility when aggregator tries to execute the frequency estimation. In this paper, a new LDP algorithm with the optimal privacy perturbation parameter based on Basic RAPPOR is proposed to improve data utility by minimizing the expected mean square error (EMSE). Then, a personalized data collection scheme based on the new LDP is elaborately presented to realize the fact that every participant can select his/her required privacy level to achieve personalized privacy preservation while guaranteeing higher data utility. Finally, the proposed personalized data collection scheme is simulated and verified on both synthetic and real datasets, which proves the feasibility and effectiveness of the proposed scheme in terms of the MSE.