Privacy-Constrained Remote Source Coding

TL;DR

This paper investigates how to efficiently compress data for sharing while ensuring privacy of secret parts, providing bounds on the tradeoffs between compression, accuracy, and privacy, and relating to the information bottleneck method.

Contribution

It formulates a new remote source coding problem with privacy constraints, deriving bounds and special cases, and connects it to the secure information bottleneck framework.

Findings

Inner and outer bounds for the rate-distortion-leakage region are established.

Bounds coincide in certain special cases, characterizing optimal tradeoffs.

Specialization to logarithmic loss relates to a secure version of the information bottleneck.

Abstract

We consider the problem of revealing/sharing data in an efficient and secure way via a compact representation. The representation should ensure reliable reconstruction of the desired features/attributes while still preserve privacy of the secret parts of the data. The problem is formulated as a remote lossy source coding with a privacy constraint where the remote source consists of public and secret parts. Inner and outer bounds for the optimal tradeoff region of compression rate, distortion, and privacy leakage rate are given and shown to coincide for some special cases. When specializing the distortion measure to a logarithmic loss function, the resulting rate-distortion-leakage tradeoff for the case of identical side information forms an optimization problem which corresponds to the "secure" version of the so-called information bottleneck.