A Seesaw Model Attack Algorithm for Distributed Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929530383368192 |
|---|---|
| author | Yang, Kun Luo, Tianyi Dong, Yanjie Li, Aohan |
| author_facet | Yang, Kun Luo, Tianyi Dong, Yanjie Li, Aohan |
| contents | We investigate the Byzantine attack problem within the context of model training in distributed learning systems. While ensuring the convergence of current model training processes, common solvers (e.g. SGD, Adam, RMSProp, etc.) can be easily compromised by malicious nodes in these systems. Consequently, the training process may either converge slowly or even diverge. To develop effective secure distributed learning solvers, it is crucial to first examine attack methods to assess the robustness of these solvers. In this work, we contribute to the design of attack strategies by initially highlighting the limitations of finite-norm attacks. We then introduce the seesaw attack, which has been demonstrated to be more effective than the finite-norm attack. Through numerical experiments, we evaluate the efficacy of the seesaw attack across various gradient aggregation rules. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_05161 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | A Seesaw Model Attack Algorithm for Distributed Learning Yang, Kun Luo, Tianyi Dong, Yanjie Li, Aohan Distributed, Parallel, and Cluster Computing We investigate the Byzantine attack problem within the context of model training in distributed learning systems. While ensuring the convergence of current model training processes, common solvers (e.g. SGD, Adam, RMSProp, etc.) can be easily compromised by malicious nodes in these systems. Consequently, the training process may either converge slowly or even diverge. To develop effective secure distributed learning solvers, it is crucial to first examine attack methods to assess the robustness of these solvers. In this work, we contribute to the design of attack strategies by initially highlighting the limitations of finite-norm attacks. We then introduce the seesaw attack, which has been demonstrated to be more effective than the finite-norm attack. Through numerical experiments, we evaluate the efficacy of the seesaw attack across various gradient aggregation rules. |
| title | A Seesaw Model Attack Algorithm for Distributed Learning |
| topic | Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2410.05161 |