[Submitted on 26 Sep 2025 (v1), last revised 15 Nov 2026 (this version, v4)]
Abstract:Federated learning (FL) enables collaborative model training among multiple clients without directly sharing their raw this http URL, gradient inversion attacks can exploit the model updates exchanged during training to reconstruct clients’ private this http URL multiple local training steps, an observed update combines gradients evaluated at changing model states and potentially different mini-batches, while intermediate states remain hidden. Restricting gradient evaluation to a single state on the segment between the initial and final models can exclude states that better fit the accumulated this http URL consider reconstruction from an individual client’s endpoint models, assuming that the training labels are this http URL propose NL-SME, a regularized two-state gradient surrogate for matching multi-step client updates without replaying local this http URL surrogate adapts its evaluation center, regularizes its displacement, and averages raw parameter gradients at two shared symmetric this http URL evaluate NL-SME against existing gradient inversion methods on FEMNIST and CIFAR-100 with varying numbers of local training this http URL results show improved reconstruction quality, highlighting the privacy risks of sharing individual client updates after multiple local steps.
Submission history
From: Li Xia [view email]
[v1] Fri, 26 Sep 2025 09:04:25 UTC (5,861 KB)
[v2] Mon, 15 Dec 2025 06:52:47 UTC (8,670 KB)
[v3] Sun, 5 Jul 2026 17:28:28 UTC (6,136 KB)
[v4] Sun, 15 Nov 2026 15:02:47 UTC (5,259 KB)
