Dynamic Byzantine-Robust Learning: Adapting to Switching Byzantine Workers

Ron Dorfman, Naseem Amin Yehya, Kfir Yehuda Levy
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:11501-11543, 2024.

Abstract

Byzantine-robust learning has emerged as a prominent fault-tolerant distributed machine learning framework. However, most techniques focus on the static setting, wherein the identity of Byzantine workers remains unchanged throughout the learning process. This assumption fails to capture real-world dynamic Byzantine behaviors, which may include intermittent malfunctions or targeted, time-limited attacks. Addressing this limitation, we propose DynaBRO – a new method capable of withstanding any sub-linear number of identity changes across rounds. Specifically, when the number of such changes is $\mathcal{O}(\sqrt{T})$ (where $T$ is the total number of training rounds), DynaBRO nearly matches the state-of-the-art asymptotic convergence rate of the static setting. Our method utilizes a multi-level Monte Carlo (MLMC) gradient estimation technique applied at the server to robustly aggregated worker updates. By additionally leveraging an adaptive learning rate, we circumvent the need for prior knowledge of the fraction of Byzantine workers.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-dorfman24a, title = {Dynamic {B}yzantine-Robust Learning: Adapting to Switching {B}yzantine Workers}, author = {Dorfman, Ron and Yehya, Naseem Amin and Levy, Kfir Yehuda}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {11501--11543}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/dorfman24a/dorfman24a.pdf}, url = {https://proceedings.mlr.press/v235/dorfman24a.html}, abstract = {Byzantine-robust learning has emerged as a prominent fault-tolerant distributed machine learning framework. However, most techniques focus on the static setting, wherein the identity of Byzantine workers remains unchanged throughout the learning process. This assumption fails to capture real-world dynamic Byzantine behaviors, which may include intermittent malfunctions or targeted, time-limited attacks. Addressing this limitation, we propose DynaBRO – a new method capable of withstanding any sub-linear number of identity changes across rounds. Specifically, when the number of such changes is $\mathcal{O}(\sqrt{T})$ (where $T$ is the total number of training rounds), DynaBRO nearly matches the state-of-the-art asymptotic convergence rate of the static setting. Our method utilizes a multi-level Monte Carlo (MLMC) gradient estimation technique applied at the server to robustly aggregated worker updates. By additionally leveraging an adaptive learning rate, we circumvent the need for prior knowledge of the fraction of Byzantine workers.} }
Endnote
%0 Conference Paper %T Dynamic Byzantine-Robust Learning: Adapting to Switching Byzantine Workers %A Ron Dorfman %A Naseem Amin Yehya %A Kfir Yehuda Levy %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-dorfman24a %I PMLR %P 11501--11543 %U https://proceedings.mlr.press/v235/dorfman24a.html %V 235 %X Byzantine-robust learning has emerged as a prominent fault-tolerant distributed machine learning framework. However, most techniques focus on the static setting, wherein the identity of Byzantine workers remains unchanged throughout the learning process. This assumption fails to capture real-world dynamic Byzantine behaviors, which may include intermittent malfunctions or targeted, time-limited attacks. Addressing this limitation, we propose DynaBRO – a new method capable of withstanding any sub-linear number of identity changes across rounds. Specifically, when the number of such changes is $\mathcal{O}(\sqrt{T})$ (where $T$ is the total number of training rounds), DynaBRO nearly matches the state-of-the-art asymptotic convergence rate of the static setting. Our method utilizes a multi-level Monte Carlo (MLMC) gradient estimation technique applied at the server to robustly aggregated worker updates. By additionally leveraging an adaptive learning rate, we circumvent the need for prior knowledge of the fraction of Byzantine workers.
APA
Dorfman, R., Yehya, N.A. & Levy, K.Y.. (2024). Dynamic Byzantine-Robust Learning: Adapting to Switching Byzantine Workers. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:11501-11543 Available from https://proceedings.mlr.press/v235/dorfman24a.html.

Related Material