ERIS：通过联邦分片聚合优化联邦学习隐私-效用-可扩展性权衡

View PDF HTML (experimental)

Abstract:Scaling Federated Learning (FL) to billion-parameter models forces a challenging trade-off between privacy, scalability, and model utility. Existing solutions often tackle these challenges in isolation, sacrificing accuracy, relying on costly cryptographic tools, or introducing communication and optimization inefficiencies that affect convergence. We introduce ERIS, an FL framework centered on Federated Shard Aggregation (FSA), a novel mechanism that partitions each client update into non-overlapping shards whose aggregation is distributed across multiple client-side aggregators. FSA removes the central aggregation bottleneck, limits the information visible to any single observer, and preserves the centralized FL update after reassembly. ERIS can further readily integrate Distributed Shifted Compression (DSC) to reduce transmitted payloads and exposed coordinates. We prove that ERIS preserves convergence under standard assumptions and bounds mutual information leakage by the observable fraction of each update, decreasing with the number of client-side aggregators, and with the compression level when DSC is enabled. Experiments across image and text tasks, including large language models, show that ERIS achieves FedAvg-level utility while substantially reducing communication bottlenecks and improving robustness to membership inference and reconstruction attacks, without relying on heavy cryptography or utility-degrading perturbations.

Comments:	Accepted at NeurIPS 2026. 53 pages
Subjects:	Machine Learning (cs.LG)
Cite as:	arXiv:2602.08617 [cs.LG]
	(or arXiv:2602.08617v3 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2602.08617 arXiv-issued DOI via DataCite

Submission history

From: Dario Fenoglio [view email]
[v1] Mon, 9 Feb 2026 13:05:41 UTC (4,676 KB)
[v2] Mon, 11 May 2026 13:31:02 UTC (6,264 KB)
[v3] Thu, 8 Oct 2026 18:45:18 UTC (6,334 KB)