• Home
  • Search
  • Defending Against Sophisticated Poisoning Attacks with RL-based Aggregation in Federated Learning
  • Cite Icon3
  • https://doi.org/10.1609/aaai.v39i24.34733Copy DOI Icon

Defending Against Sophisticated Poisoning Attacks with RL-based Aggregation in Federated Learning

  • Abstract
  • Literature Map
  • Citations
  • Similar Papers
Abstract

Federated learning is susceptible to model poisoning attacks, especially those meticulously crafted for servers. Traditional defense methods mainly focus on updating assessments or robust aggregation against manually crafted myopic attacks. When facing advanced attacks, their defense stability is notably insufficient. Therefore, it is imperative to develop adaptive defenses against such advanced poisoning attacks. We find that benign clients exhibit significantly higher data distribution stability than malicious clients in federated learning in both CV and NLP tasks. Therefore, the malicious clients can be recognized by observing the stability of their data distribution. In this paper, we propose AdaAggRL, an RL-based Adaptive Aggregation method, to defend against sophisticated poisoning attacks. Specifically, we first utilize distribution learning to simulate the clients' data distributions. Then, we use maximum mean discrepancy (MMD) to calculate the pairwise similarity of the current local model data distribution, its historical data distribution, and global model data distribution. Finally, we use policy learning to adaptively determine the aggregation weights based on the above similarities. Experiments on four real-world datasets demonstrate that the proposed defense model significantly outperforms widely adopted defense models for sophisticated attacks.

Similar Papers
  • Research Article
  • Citations2

Attacks and countermeasures on federated learning via historical knowledge modeling

  • Jul 17, 2025
  • Journal of King Saud University Computer and Information Sciences
  • Songsong Zhang +4
  • Research Article
  • Citations19

Center-aligned domain adaptation network for image classification

  • Nov 27, 2020
  • Expert Systems with Applications
  • Guanqun Wei +4
  • Research Article
  • Citations94

Citizen-Based Litter and Marine Debris Data Collection and Mapping

  • Jul 01, 2015
  • Computing in Science & Engineering
  • Jenna R Jambeck +1
  • Research Article
  • Citations7

A cross-dataset adaptive domain selection transfer learning framework for motor imagery-based brain-computer interfaces

  • Jun 01, 2024
  • Journal of Neural Engineering
  • Jing Jin +6
  • Dissertation

Learning Invariances for High-Dimensional Data Analysis

  • Jan 01, 2014
  • The University of Queensland
  • Mahsa Baktashmotlagh
  • Research Article

Research on Transfer Learning-Based Fault Diagnosis for Planetary Gearboxes Under Cross-Operating Conditions via IDANN

  • Dec 18, 2025
  • Information
  • Xiaolu Wang +3
  • Research Article
  • Citations1

Different estimation methods of the modified Kies Topp-Leone model with applications and quantile regression.

  • Sep 13, 2024
  • PloS one
  • Safar M Alghamdi +5
  • PDF
  • Research Article
  • Citations7

Nuclear discrepancy for single-shot batch active learning

  • Jun 26, 2019
  • Machine Learning
  • Tom J Viering +2
  • PDF
  • Research Article
  • Citations6

Understanding Collections of Related Datasets Using Dependent MMD Coresets

  • Sep 23, 2021
  • Information
  • Sinead A Williamson +1
  • Research Article
  • Citations143

Distributionally Adversarial Attack

  • Jul 17, 2019
  • Proceedings of the AAAI Conference on Artificial Intelligence
  • Tianhang Zheng +2
  • Book Chapter
  • Citations6

Improving Maximum Classifier Discrepancy by Considering Joint Distribution for Domain Adaptation

  • Jan 01, 2018
  • Zehang Lin +6
  • Research Article
  • Citations11

Hierarchical federated learning with local model embedding

  • Apr 01, 2023
  • Engineering Applications of Artificial Intelligence
  • Yunlong He +2
  • Research Article

Projection Kernel regularization for diffusion-based multimodal remote sensing segmentation.

  • Mar 21, 2026
  • Scientific reports
  • Xu Tong +7
  • Research Article
  • Citations10

Interactive attack-defense for generalized person re-identification

  • May 03, 2024
  • Neural Networks
  • Huafeng Li +4
  • Research Article

Research on cross-domain bolt looseness detection of cylindrical shell structures based on a distorted similitude model

  • Oct 07, 2025
  • Structural Health Monitoring
  • Pengtao Liu +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.