AI Paper Atlas

Research map / Federated learning and privacy / Privacy attacks and secure learning

Robust Federated Inference

Akash Dhasade, Sadegh Farhadkhani, Rachid Guerraoui, Nirupam Gupta, Maxime Jacovella, Anne-Marie Kermarrec, Rafael Pinot

ICLR 2026 · Federated learning and privacy › Privacy attacks and secure learning · approach: Defending federated learning against poisoning · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use adversarially trained DeepSet with test-time trimmed mean (DeepSet-TM) instead of best static robust aggregatorDeepSet-TM · CIFAR-10, CIFAR-100, AG-Newsworst-case accuracy: +4.7 to +22.2 pointsImproved
use DeepSet-TM instead of static aggregation under PGD-cwDeepSet-TM · AG-Newsaccuracy: 55.3 -> 83.2Improved
use DeepSet-TM instead of CWTM under SIA white-boxDeepSet-TM · CIFAR-10accuracy: 44.8 -> 51.4Improved
use DeepSet-TM instead of Mean under logit flippingDeepSet-TM · CIFAR-100accuracy: 78.8 -> 78.0No clear effect
scale clients to n=25, f=5DeepSet-TM · CIFAR-100 (alpha=0.3)worst-case accuracy: 30.4 -> 50.7 (vs CWTM)Improved
use DeepSet-TM instead of CoPur under logit flipping with larger amplification or more adversariesDeepSet-TM · CIFAR-10accuracyImproved
add adversarial training to DeepSet aggregator (ablation)DeepSet-TM · no-adversary case (f=0)clean accuracyHurt
use CWMed instead of MeanCWMed · CIFAR-100 (alpha=0.5)accuracy under logit flipping: 78.8 -> 65.8Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy