Estimation Efficiency Under Privacy Constraints

DOI10.1109/TIT.2018.2865558MaRDI QIDQ4629912zbMATH OpenOpenAlexWikidataFDO

Authors Shahab Asoodeh, M. C. Díaz, Fady Alajaji, Tamás Linder

Publication date 28 March 2019

Published in IEEE Transactions on Information Theory (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1707.02409

Mathematics Subject Classification ID

Measures of information, entropy (94A17) Source coding (94A29) Authentication, digital signatures and secret sharing (94A62)

Abstract: We investigate the problem of estimating a random variable $YinmathcalY$ under a privacy constraint dictated by another random variable $XinmathcalX$, where estimation efficiency and privacy are assessed in terms of two different loss functions. In the discrete case, we use the Hamming loss function and express the corresponding utility-privacy tradeoff in terms of the privacy-constrained guessing probability $h(P_{XY},epsilon)$, the maximum probability $mathsf{P}_{m}athsfc(Y|Z)$ of correctly guessing $Y$ given an auxiliary random variable $ZinmathcalZ$, where the maximization is taken over all $P_{Z|Y}$ ensuring that $mathsf{P}_{m}athsfc(X|Z)leqepsilon$ for a given privacy threshold $epsilongeq0$. We prove that $h(P_{XY},cdot)$ is concave and piecewise linear, which allows us to derive its expression in closed form for any $epsilon$ when $X$ and $Y$ are binary. In the non-binary case, we derive $h(P_{XY},epsilon)$ in the high utility regime (i.e., for sufficiently large values of $epsilon$) under the assumption that $Z$ takes values in $mathcalY$. We also analyze the privacy-constrained guessing probability for two binary vector scenarios. When $X$ and $Y$ are continuous random variables, we use the squared-error loss function and express the corresponding utility-privacy tradeoff in terms of $mathsfsENSR(P_{XY},epsilon)$, which is the smallest normalized minimum mean squared-error (mmse) incurred in estimating $Y$ from its Gaussian perturbation $Z$, such that the mmse of $f(X)$ given $Z$ is within $epsilon$ of the variance of $f(X)$ for any non-constant real-valued function $f$. We derive tight upper and lower bounds for $mathsfsENSR$ when $Y$ is Gaussian. We also obtain a tight lower bound for $mathsfsENSR(P_{XY},epsilon)$ for general absolutely continuous random variables when $epsilon$ is sufficiently small.

Cited in

(6)

• Private versus verifiable interim performance evaluations under uncertainty
• Optimal Privacy-Aware Estimation
• Secure Estimation Under Causative Attacks
• PRIVACY RISK ASSESSMENT WITH BOUNDS DEDUCED FROM BOUNDS
• The cost of privacy: optimal rates of convergence for parameter estimation with differential privacy
• Comparison of contraction coefficients for f-divergences

This page was built for publication: Estimation Efficiency Under Privacy Constraints

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4629912)