Large Margin Multiclass Gaussian Classification with Differential Privacy

TL;DR

This paper introduces a differentially private multi-class Gaussian classifier using a large margin loss, providing theoretical bounds on the excess risk caused by privacy-preserving perturbations.

Contribution

It proposes a novel differentially private algorithm for multiclass Gaussian classification with a large margin loss, including theoretical risk bounds.

Findings

The algorithm satisfies differential privacy constraints.

Theoretical upper bound on excess risk due to privacy perturbation.

Effective classification performance under privacy guarantees.

Abstract

As increasing amounts of sensitive personal information is aggregated into data repositories, it has become important to develop mechanisms for processing the data without revealing information about individual data instances. The differential privacy model provides a framework for the development and theoretical analysis of such mechanisms. In this paper, we propose an algorithm for learning a discriminatively trained multi-class Gaussian classifier that satisfies differential privacy using a large margin loss function with a perturbed regularization term. We present a theoretical upper bound on the excess risk of the classifier introduced by the perturbation.