Improved Generalization Guarantees in Restricted Data Models

TL;DR

This paper demonstrates that in data models with weakly correlated distant attributes, differential privacy can be more effectively reused across data segments, leading to improved accuracy and generalization guarantees.

Contribution

It introduces a novel approach to reusing privacy budget in weakly correlated data models, enhancing privacy-utility trade-offs in differential privacy.

Findings

Reusing privacy budget improves accuracy in weakly correlated data models.

Weak correlations enable better privacy-utility trade-offs.

Enhanced generalization guarantees under the proposed data model.

Abstract

Differential privacy is known to protect against threats to validity incurred due to adaptive, or exploratory, data analysis -- even when the analyst adversarially searches for a statistical estimate that diverges from the true value of the quantity of interest on the underlying population. The cost of this protection is the accuracy loss incurred by differential privacy. In this work, inspired by standard models in the genomics literature, we consider data models in which individuals are represented by a sequence of attributes with the property that where distant attributes are only weakly correlated. We show that, under this assumption, it is possible to "re-use" privacy budget on different portions of the data, significantly improving accuracy without increasing the risk of overfitting.