Robust Federated Learning by Mixture of Experts

Saeedeh Parsaeefard, Sayed Ehsan Etesami, Alberto Leon Garcia
4/23/2021
cs.LGcs.DCcs.NE
moefederated

Official code: etesami/MOE-FL (4 stars, Apache-2.0, last commit 2021-05-09).

Abstract

We present a novel weighted average model based on the mixture of experts (MoE) concept to provide robustness in Federated learning (FL) against the poisoned/corrupted/outdated local models. These threats along with the non-IID nature of data sets can considerably diminish the accuracy of the FL model. Our proposed MoE-FL setup relies on the trust between users and the server where the users share a portion of their public data sets with the server. The server applies a robust aggregation method by solving the optimization problem or the Softmax method to highlight the outlier cases and to reduce their adverse effect on the FL process. Our experiments illustrate that MoE-FL outperforms the performance of the traditional aggregation approach for high rate of poisoned data from attackers.

View on arXivView PDF

Code Implementations(1)

etesami/MOE-FLOfficial100%
Apache-2.0

The implementation of the "Robust Federated Learning by Mixture of Experts" study.

41Shell, PythonMay 9, 20205 years agoApache-2.0
federated-learningmixture-of-expertsprivacy-preservingpysyft

Similar Papers

Suppressing Poisoning Attacks on Federated Learning for Medical Imaging

84% match1 repo72022

FreqFed: A Frequency Analysis-Based Approach for Mitigating Poisoning Attacks in Federated Learning

83% match2023

Robust Aggregation for Federated Learning

83% match1 repo532019

Sentinel: An Aggregation Function to Secure Decentralized Federated Learning

82% match2023

AGRAMPLIFIER: Defending Federated Learning Against Poisoning Attacks Through Local Update Amplification

82% match2023

Celtibero: Robust Layered Aggregation for Federated Learning

81% match2024

Cite this paper

@article{parsaeefard2021robust,
  title  = {Robust Federated Learning by Mixture of Experts},
  author = {Saeedeh Parsaeefard and Sayed Ehsan Etesami and Alberto Leon Garcia},
  year   = {2021},
  eprint = {2104.11700},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2104.11700}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2104.11700.svg)](https://codeofpaper.com/paper/2104.11700)

Discussion