Constructing Privacy Channels from Information Channels

TL;DR

This paper introduces a method to construct privacy channels using information theory, enabling better privacy protection by evaluating and limiting the information disclosed about individuals in data queries.

Contribution

It transforms the complex problem of assessing infinite channel capacities into a finite problem, facilitating the construction of various privacy channels.

Findings

Provides a finite-capacity evaluation method for privacy channels

Constructs basic privacy channels like random response, exponential, and Gaussian channels

Links privacy channels to differential privacy mechanisms

Abstract

Data privacy protection studies how to query a dataset while preserving the privacy of individuals whose sensitive information is contained in the dataset. The information privacy model protects the privacy of an individual by using a noisy channel, called privacy channel, to filter out most information of the individual from the query's output. This paper studies how to construct privacy channels, which is challenging since it needs to evaluate the maximal amount of disclosed information of each individual contained in the query's output, called individual channel capacity. Our main contribution is an interesting result which can transform the problem of evaluating a privacy channel's individual channel capacity, which equals the problem of evaluating the capacities of an infinite number of channels, into the problem of evaluating the capacities of a finite number of channels. This result gives us a way to utilize the results in the information theory to construct privacy channels. As some examples, it is used to construct several basic privacy channels, such as the random response privacy channel, the exponential privacy channel and the Gaussian privacy channel, which are respective counterparts of the random response mechanism, the exponential mechanism and the Gaussian mechanism of differential privacy.