14
3

Phase transitions for support recovery under local differential privacy

Abstract

We address the problem of variable selection in a high-dimensional but sparse mean model, under the additional constraint that only privatised data are available for inference. The original data are vectors with independent entries having a symmetric, strongly log-concave distribution on R\mathbb{R}. For this purpose, we adopt a recent generalisation of classical minimax theory to the framework of local α\alpha-differential privacy. We provide lower and upper bounds on the rate of convergence for the expected Hamming loss over classes of at most ss-sparse vectors whose non-zero coordinates are separated from 00 by a constant a>0a>0. As corollaries, we derive necessary and sufficient conditions (up to log factors) for exact recovery and for almost full recovery. When we restrict our attention to non-interactive mechanisms that act independently on each coordinate our lower bound shows that, contrary to the non-private setting, both exact and almost full recovery are impossible whatever the value of aa in the high-dimensional regime such that nα2/d21n \alpha^2/ d^2\lesssim 1. However, in the regime nα2/d2log(d)n\alpha^2/d^2\gg \log(d) we can exhibit a critical value aa^* (up to a logarithmic factor) such that exact and almost full recovery are possible for all aaa\gg a^* and impossible for aaa\leq a^*. We show that these results can be improved when allowing for all non-interactive (that act globally on all coordinates) locally α\alpha-differentially private mechanisms in the sense that phase transitions occur at lower levels.

View on arXiv
Comments on this paper