Privacy-Utility Tradeoff Based on $\alpha$-lift

TL;DR

This paper explores the privacy-utility tradeoff using $eta$-lift, a tunable privacy measure, proposing a heuristic algorithm to optimize utility across different $eta$ values and demonstrating its effectiveness through numerical experiments.

Contribution

It introduces a heuristic algorithm for optimizing the privacy-utility tradeoff with $eta$-lift, addressing the nonlinear challenges for finite $eta$, and proves the convexity of $eta$-lift.

Findings

The algorithm effectively estimates optimal utility for various $eta$ values.

Numerical results validate the algorithm's efficacy and reveal the effective range of $eta$ and privacy budget.

The convexity of $eta$-lift with respect to lift is established.

Abstract

Information density and its exponential form, known as lift, play a central role in information privacy leakage measures. $\alpha$-lift is the power-mean of lift, which is tunable between the worst-case measure max-lift ($\alpha=\infty$) and more relaxed versions ($\alpha<\infty$). This paper investigates the optimization problem of the privacy-utility tradeoff (PUT) where $\alpha$-lift and mutual information are privacy and utility measures, respectively. Due to the nonlinear nature of $\alpha$-lift for $\alpha<\infty$, finding the optimal solution is challenging. Therefore, we propose a heuristic algorithm to estimate the optimal utility for each value of $\alpha$, inspired by the optimal solution for $\alpha=\infty$ and the convexity of $\alpha$-lift with respect to the lift, which we prove. The numerical results show the efficacy of the algorithm and indicate the effective range of $\alpha$ and privacy budget $\varepsilon$ with good PUT performance.