Private Algorithms Can Always Be Extended

Abstract

We consider the following fundamental question on $\epsilon$-differential privacy. Consider an arbitrary $\epsilon$-differentially private algorithm defined on a subset of the input space. Is it possible to extend it to an $\epsilon'$-differentially private algorithm on the whole input space for some $\epsilon'$ comparable with $\epsilon$? In this note we answer affirmatively this question for $\epsilon'=2\epsilon$. Our result applies to every input metric space and space of possible outputs. This result originally appeared in a recent paper by the authors [BCSZ18]. We present a self-contained version in this note, in the hopes that it will be broadly useful.