Graphical-model based estimation and inference for differential privacy

About

Many privacy mechanisms reveal high-level information about a data distribution through noisy measurements. It is common to use this information to estimate the answers to new queries. In this work, we provide an approach to solve this estimation problem efficiently using graphical models, which is particularly effective when the distribution is high-dimensional but the measurements are over low-dimensional marginals. We show that our approach is far more efficient than existing estimation techniques from the privacy literature and that it can improve the accuracy and scalability of many state-of-the-art mechanisms.

Ryan McKenna, Daniel Sheldon, Gerome Miklau · 2019

Related benchmarks

Task	Dataset	Result
Offline Reinforcement Learning	MuJoCo HalfCheetah	Normalized Return4.5	97
Offline Reinforcement Learning	Kitchen Partial	Normalized Score3.5	69
Offline Reinforcement Learning	Maze2D medium	Normalized Return46.8	38
Offline Reinforcement Learning	Maze2D umaze	Normalized Return41.6	38
Offline Reinforcement Learning	Maze2D large	Normalized Return21.9	33
Transition Synthesis	Maze2D umaze	Marginal79.3	5
Transition Synthesis	Maze2D medium	Marginal80.1	5
Transition Synthesis	Kitchen Partial	Marginal78.3	5
Transition Synthesis	Maze2D large	Marginal0.803	5
Transition Synthesis	MuJoCo HalfCheetah	Marginal0.776	5