A Privacy-Preserving User-Centric Data-Sharing Scheme

Abstract

Using raw sensitive data of end-users helps service providers manage their operations efficiently and provide high-quality services to end-users. Although access to sensitive information benefits both parties, it poses several challenges concerning end-user privacy. Most data-sharing schemes based on differential privacy allow control of the level of privacy, which is not straightforward for end-users and leads to unpredictable utility. To address this issue, a novel local differentially private data-sharing scheme is proposed featuring a bimodal probability distribution that allows determining the range of random variables from which the noise is drawn with high probability. Additionally, a local differentially private mechanism is introduced to regulate the amount of noise injected into the data to control data utility. These components are combined to make up a user-centric data-sharing scheme which provides the end-user with control over the utility of their data, with the level of privacy being calculated from individual utility preferences. The simulation results show that the proposed scheme allows keeping the utility within the boundaries defined by the end-user, while providing the maximum possible level of privacy. Furthermore, it allows injecting more noise into the data for the same error in utility compared to the Laplace mechanism.

Keywords

• Data utility
• local differential privacy
• personal data
• privacy-preserving
• probability distribution