Locally differential private data transmission exploiting the nature of wireless environment

Abstract

As the big data industry expands, various institutions and companies are collecting information of people for analyzing the statistics or developing personalized system. To prevent the leakage of private information from providing data, a privacy protection technique basedon the notion of local differential privacy (LDP) has been widely studied. Basically, the data should be perturbed randomly by some privacy mechanism before providing to others in order to satisfy the privacy constraint. In this thesis, we consider an user data distribution estimation task under LDP constraint in wireless communication environments where the sensitive user data is transmitted. We propose a new privacy protection mechanism using a combination of randomized response, known as optimal digital data preprocessing in the existing pure LDP constraint, and binary phase shift keying (BPSK), known as analog data preprocessing method and exploiting the wireless channel noise. Finally, we analyze the privacy‑utility trade‑off of the proposed scheme and we confirm that we achieve improved trade‑off comparing power control Gaussian mechanism in high privacy regime.