Sequential Subspace Noise Injection Prevents Accuracy Collapse in Certified Unlearning

TL;DR

The paper introduces sequential subspace noise injection for certified unlearning, significantly improving model accuracy while maintaining privacy guarantees, thus making certified unlearning more practical.

Contribution

It proposes a novel noise scheduling method that distributes noise across subspaces, enhancing accuracy without sacrificing privacy guarantees.

Findings

Subspace noise scheduling improves accuracy after unlearning.

Method maintains the same privacy guarantees as traditional approaches.

Empirical results show robustness against membership inference attacks.

Abstract

Certified unlearning based on differential privacy offers strong guarantees but remains largely impractical: the noisy fine-tuning approaches proposed so far achieve these guarantees but severely reduce model accuracy. We propose sequential noise scheduling, which distributes the noise budget across orthogonal subspaces of the parameter space, rather than injecting it all at once. This simple modification mitigates the destructive effect of noise while preserving the original certification guarantees. We extend the analysis of noisy fine-tuning to the subspace setting, proving that the same $(\varepsilon,\delta)$ privacy budget is retained. Empirical results on image classification benchmarks show that our approach substantially improves accuracy after unlearning while remaining robust to membership inference attacks. These results show that certified unlearning can achieve both rigorous guarantees and practical utility.