Sequential Subspace Noise Injection Prevents Accuracy Collapse in Certified Unlearning
This work addresses the practical utility issue in certified unlearning for machine learning models, though it is incremental as it modifies existing noisy fine-tuning approaches.
The paper tackles the problem of severe accuracy reduction in certified unlearning methods based on differential privacy by proposing sequential noise scheduling across orthogonal subspaces, which improves accuracy on image classification benchmarks while maintaining privacy guarantees.
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 $(\varepsilon,δ)$ 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.