Differentially Private Ensemble Classifiers for Data Streams

TL;DR

This paper introduces a differentially private ensemble method for data streams that supports unlimited updates, is model-agnostic, and effectively handles concept drift while preserving privacy.

Contribution

It proposes a novel ensemble approach enabling unbounded updates and model-agnostic privacy preservation in streaming data classification.

Findings

Outperforms existing methods on real-world datasets.

Effectively manages concept drift with privacy guarantees.

Supports an unbounded number of ensemble updates.

Abstract

Learning from continuous data streams via classification/regression is prevalent in many domains. Adapting to evolving data characteristics (concept drift) while protecting data owners' private information is an open challenge. We present a differentially private ensemble solution to this problem with two distinguishing features: it allows an \textit{unbounded} number of ensemble updates to deal with the potentially never-ending data streams under a fixed privacy budget, and it is \textit{model agnostic}, in that it treats any pre-trained differentially private classification/regression model as a black-box. Our method outperforms competitors on real-world and simulated datasets for varying settings of privacy, concept drift, and data distribution.