Leveraging Per-Instance Privacy for Machine Unlearning

TL;DR

This paper introduces a per-instance privacy approach for machine unlearning, improving utility-privacy tradeoffs by analyzing individual data points and demonstrating practical effectiveness through empirical results with SGLD and fine-tuning.

Contribution

It develops a novel per-instance privacy analysis framework for unlearning, linking data difficulty metrics with privacy loss, and proposes adaptive unlearning strategies.

Findings

Per-instance privacy losses correlate with data difficulty metrics.

Theoretical predictions match empirical results for SGLD and fine-tuning.

Harder data groups are identified and evaluated using new methods.

Abstract

We present a principled, per-instance approach to quantifying the difficulty of unlearning via fine-tuning. We begin by sharpening an analysis of noisy gradient descent for unlearning (Chien et al., 2024), obtaining a better utility-unlearning tradeoff by replacing worst-case privacy loss bounds with per-instance privacy losses (Thudi et al., 2024), each of which bounds the (Renyi) divergence to retraining without an individual data point. To demonstrate the practical applicability of our theory, we present empirical results showing that our theoretical predictions are born out both for Stochastic Gradient Langevin Dynamics (SGLD) as well as for standard fine-tuning without explicit noise. We further demonstrate that per-instance privacy losses correlate well with several existing data difficulty metrics, while also identifying harder groups of data points, and introduce novel evaluation methods based on loss barriers. All together, our findings provide a foundation for more efficient and adaptive unlearning strategies tailored to the unique properties of individual data points.