Unified Expression of Utility-Privacy Trade-off in Privacy-Constrained Source Coding

TL;DR

This paper develops a unified theoretical framework using rate-distortion theory to characterize the utility-privacy trade-off in privacy-constrained source coding, encompassing previous special cases and highlighting the importance of encoding choices.

Contribution

It introduces a comprehensive rate-distortion-based expression for the utility-privacy trade-off in the general case where encoded messages include both public and private information.

Findings

The unified expression includes previous special cases as subsets.

Numerical results show optimal encoding strategies vary with the scenario.

Neither of the previous specific cases always yields the best trade-off.

Abstract

Privacy-constrained source coding problems have become increasingly important recently, and the utility-privacy trade-off has been investigated for various systems. As pioneering work, Yamamoto (1983) found theoretical limits of the coding rate, privacy and utility in two cases; (i) both public and private information is encoded and (ii) only public information is encoded. However, the theoretical limit has not been characterized in a more general case; (iii) encoded messages consist of public information and a part of private information. Then in this paper, we characterize the trade-off relation in case (iii) using rate-distortion theory. The obtained expression of the achievable region is a "unified expression" because it includes the ones in cases (i) and (ii) as special cases. The numerical results also demonstrate that neither case (i) nor (ii) are the best cases, and it is important to select the encoded information adequately in case (iii).