Protection Against Reconstruction and Its Applications in Private Federated Learning

TL;DR

This paper develops new locally differentially private mechanisms for large-scale statistical learning, enabling practical private model training with minimal utility loss across various applications.

Contribution

It introduces a reconceptualization of privacy protections for adversaries with limited prior information, allowing larger privacy parameters and designing minimax optimal mechanisms.

Findings

Able to train large-scale image and language models privately

Achieves minimal utility degradation at high privacy levels

Provides both theoretical guarantees and empirical validation

Abstract

In large-scale statistical learning, data collection and model fitting are moving increasingly toward peripheral devices---phones, watches, fitness trackers---away from centralized data collection. Concomitant with this rise in decentralized data are increasing challenges of maintaining privacy while allowing enough information to fit accurate, useful statistical models. This motivates local notions of privacy---most significantly, local differential privacy, which provides strong protections against sensitive data disclosures---where data is obfuscated before a statistician or learner can even observe it, providing strong protections to individuals' data. Yet local privacy as traditionally employed may prove too stringent for practical use, especially in modern high-dimensional statistical and machine learning problems. Consequently, we revisit the types of disclosures and adversaries against which we provide protections, considering adversaries with limited prior information and ensuring that with high probability, ensuring they cannot reconstruct an individual's data within useful tolerances. By reconceptualizing these protections, we allow more useful data release---large privacy parameters in local differential privacy---and we design new (minimax) optimal locally differentially private mechanisms for statistical learning problems for \emph{all} privacy levels. We thus present practicable approaches to large-scale locally private model training that were previously impossible, showing theoretically and empirically that we can fit large-scale image classification and language models with little degradation in utility.