Byzantine-Robust Gossip: Insights from a Dual Approach
Abstract
Distributed learning has many computational benefits but is vulnerable to attacks from a subset of devices transmitting incorrect information. This paper investigates Byzantine-resilient algorithms in a decentralized setting, where devices communicate directly in a peer-to-peer manner within a communication network. We leverage the so-called dual approach for decentralized optimization and propose a Byzantine-robust algorithm. We provide convergence guarantees in the average consensus subcase, discuss the potential of the dual approach beyond this subcase, and re-interpret existing algorithms using the dual framework. Lastly, we experimentally show the soundness of our method.
Cite
Text
Gaucher et al. "Byzantine-Robust Gossip: Insights from a Dual Approach." Transactions on Machine Learning Research, 2026.Markdown
[Gaucher et al. "Byzantine-Robust Gossip: Insights from a Dual Approach." Transactions on Machine Learning Research, 2026.](https://mlanthology.org/tmlr/2026/gaucher2026tmlr-byzantinerobust/)BibTeX
@article{gaucher2026tmlr-byzantinerobust,
title = {{Byzantine-Robust Gossip: Insights from a Dual Approach}},
author = {Gaucher, Renaud and Hendrikx, Hadrien and Dieuleveut, Aymeric},
journal = {Transactions on Machine Learning Research},
year = {2026},
url = {https://mlanthology.org/tmlr/2026/gaucher2026tmlr-byzantinerobust/}
}