DP-{\lambda}CGD: Efficient Noise Correlation for Differentially Private Model Training

TL;DR

This paper introduces DP-λCGD, a noise correlation method for DP-SGD that improves accuracy and reduces memory overhead by correlating noise only with the previous iteration and regenerating it via pseudorandom generators.

Contribution

It proposes a novel noise correlation strategy that minimizes memory use and computational overhead while enhancing privacy-preserving model training accuracy.

Findings

Empirically improved accuracy over standard DP-SGD.

Requires no additional memory beyond standard DP-SGD.

Minimal computational overhead introduced.

Abstract

Differentially private stochastic gradient descent (DP-SGD) is the gold standard for training machine learning models with formal differential privacy guarantees. Several recent extensions improve its accuracy by introducing correlated noise across training iterations. Matrix factorization mechanisms are a prominent example, but they correlate noise across many iterations and require storing previously added noise vectors, leading to substantial memory overhead in some settings. In this work, we propose a new noise correlation strategy that correlates noise only with the immediately preceding iteration and cancels a controlled portion of it. Our method relies on noise regeneration using a pseudorandom noise generator, eliminating the need to store past noise. As a result, it requires no additional memory beyond standard DP-SGD. We show that the computational overhead is minimal and empirically demonstrate improved accuracy over DP-SGD.