Differentially Private Manifold Denoising
arXiv:2604.00942v1 Announce Type: new
Abstract: We introduce a differentially private manifold denoising framework that allows users to exploit sensitive reference datasets to correct noisy, non-private query points without compromising privacy. The m…