pith. sign in

Optimal error of query sets under the differentially-private matrix mechanism

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
abstract

A common goal of privacy research is to release synthetic data that satisfies a formal privacy guarantee and can be used by an analyst in place of the original data. To achieve reasonable accuracy, a synthetic data set must be tuned to support a specified set of queries accurately, sacrificing fidelity for other queries. This work considers methods for producing synthetic data under differential privacy and investigates what makes a set of queries "easy" or "hard" to answer. We consider answering sets of linear counting queries using the matrix mechanism, a recent differentially-private mechanism that can reduce error by adding complex correlated noise adapted to a specified workload. Our main result is a novel lower bound on the minimum total error required to simultaneously release answers to a set of workload queries. The bound reveals that the hardness of a query workload is related to the spectral properties of the workload when it is represented in matrix form. The bound is most informative for $(\epsilon,\delta)$-differential privacy but also applies to $\epsilon$-differential privacy.

fields

cs.LG 1

years

2019 1

verdicts

UNVERDICTED 1

representative citing papers

Capacity Bounded Differential Privacy

cs.LG · 2019-07-03 · unverdicted · novelty 7.0

Defines capacity-bounded differential privacy via restricted f-divergences to model adversaries limited by function class capacity.

citing papers explorer

Showing 1 of 1 citing paper.

  • Capacity Bounded Differential Privacy cs.LG · 2019-07-03 · unverdicted · none · ref 19 · internal anchor

    Defines capacity-bounded differential privacy via restricted f-divergences to model adversaries limited by function class capacity.