The matrix mechanism: optimizing linear counting queries under differential privacy

By Chao Li, Gerome Miklau, Michael Hay, Andrew McGregor, Vibhor Rastogi. 2015

206 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

An Economic Analysis of Privacy Protection and Statistical Accuracy as Social Choices (2018, background): “(2006) proved the single query version of the Laplace mechanism, which Li et al. (2015) generalized to a matrix workload.”
Principled Evaluation of Differentially Private Algorithms using DPBench (2015, methodology): “It can be shown that all of the data independent algorithms studied here are instances of the matrix mechanism [16, 18], a generic framework that computes linear combinations of cell counts, adds noise, and then reconstructs estimates for the individual cells by linear transformations.”
Differentially Private Data Publishing and Analysis: A Survey (2017, background): “The error bound of the matrix mechanism was analyzed in their subsequent paper [38].”
Optimizing error of high-dimensional statistical queries under differential privacy (2018, methodology): “Strategy A supports a workload W if and only if every query in W can be expressed as a linear combination of the queries in A, which occurs whenever W = WAA [28].”
EKTELO: A Framework for Defining Differentially-Private Computations (2018, methodology): “The use of inference appears in many di↵erentially private algorithms [3, 5, 7, 18, 22, 23, 25, 31, 36, 37, 43].”
PrivateSQL: A Differentially Private SQL Query Engine (2019, methodology): “More specifically, we use: W-nnls, a workload-aware version of non-negative least squares inference [28], and the workloadagnostic algorithms Identity and Part, the latter of which performs the partitioning step of the DAWA algorithm [27].”