Predictable Privacy-Preserving Mobile Crowd Sensing: A Tale of Two Roles

Abstract

The rise of mobile crowd sensing has brought privacy issues into a sharp view. In this paper, our goal is to achieve the predictable privacy-preserving mobile crowd sensing, which we envision to have the capability to quantify the privacy protections, and simultaneously allowing application users to predict the utility loss at the same time. The Salus algorithm is first proposed to protect the private data against the data reconstruction attacks. To understand privacy protection, we quantify the privacy risks in terms of private data leakage under reconstruction attacks. To predict the utility, we provide accurate utility predictions for various crowd sensing applications using Salus. The risk assessments can be generally applied to different type of sensors on the mobile platform, and the utility prediction can also be used to support various applications that use data aggregators such as average, histogram, and classifiers. Finally, we propose and implement the $P^{3}$ application framework. Both measurement results using online datasets and real-world case studies show that the $P^{3}$ provides accurate risk assessments and utility estimations, which makes it a promising framework to support future privacy-preserving mobilecrowd sensing applications.