AI Paper Atlas

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

Towards Attack-tolerant Federated Learning via Critical Parameter Analysis

Sungwon Han, Sungwon Park, Fangzhao Wu, Sundong Kim, Bin Zhu, Xing Xie, Meeyoung Cha

ICCV 2023 · Federated learning and privacy › Privacy attacks and secure learning · approach: Defending federated learning against poisoning · 28 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
add bottom-k parameter comparison to the normality score (ablation)FedCPA · CIFAR-10 targeted attackattack success rate (%): 36.8 -> 12.5Improved
add top-k parameter comparison to the normality score (ablation)FedCPA · CIFAR-10 targeted attackattack success rate (%): 18.9 -> 12.5Improved
add similarity to previous-round global model (ablation)FedCPA · CIFAR-10 targeted attackattack success rate (%): 24.1 -> 12.5Improved
add similarity among local models (ablation)FedCPA · CIFAR-10 targeted attackattack success rate (%): 20.2 -> 12.5Improved
set k to 1% rather than 0.5% or 5% (ablation)FedCPA · CIFAR-10 targeted attackattack success rate (%): 63.4 (0.5%) / 51.6 (5%) -> 12.5 (1%)Improved
use FedCPA instead of baseline aggregators under targeted attackFedAvg · CIFAR-10 targeted (pollution 0.5)attack success rate (%): 51.4 -> 21.9Improved
use FedCPA instead of baselines on clean accuracy under TinyImageNet targeted attackFedAvg · TinyImageNet targeted (pollution 0.5)accuracy (%): 39.6 -> 30.1Hurt
use FedCPA instead of ResidualBase under Gaussian noise attackFedAvg · TinyImageNet Gaussian attackaccuracy (%): 37.0 -> 36.1No clear effect

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