A privacy-preserving data aggregation of mobile crowdsensing based on local differential privacy

Abstract

Mobile crowdsensing (MCS) is increasingly being used in smart city research to collect data, such as environmental assessment and traffic monitoring. However, this approach introduces a number of privacy and efficiency challenges, as sensing report includes the user's sensitive location and assigned attributes. Many methods adopt differential privacy scheme to protect users' privacy, while the assumption that the server is trusted is not realistic in practical application. Recently, local differential privacy has paved the way for more efficient and private data collection for the untrusted model, though it remains a challenge to obtain effective statistical analysis when applied to small and medium-sized MCS tasks. In this paper, we improve the local ∈-differential privacy method for MCS data aggregation to preserve participant privacy and achieve accurate data analysis. Considering the different attributes of sensing data, we first adopt a distinct local differential privacy procedure to diverse sensing attributes. Then we propose a data aggregation algorithm to count and remove the noise data provided by participants. Simulation results show that the proposed scheme improves analysis accuracy and reduces the lowest number of participants in a task, compared with existing similar solutions.