C OORDINATED A TTACKS A GAINST F EDERATED L EARNING : A M ULTI -A GENT R EINFORCEMENT L EARNING A PPROACH
C OORDINATED A TTACKS A GAINST F EDERATED L EARNING : A M ULTI -A GENT R EINFORCEMENT L EARNING A PPROACH
复制标题
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Wen Shen;Henger Li;Zizhan Zheng
中科院分区：
文献类型：
--
作者：
Wen Shen;Henger Li;Zizhan Zheng
我们提出了一个基于模型的强化学习框架对联邦学习系统。我们的方法首先通过合作的多智能体协调来近似聚合数据的分布。然后，它通过多智能体强化学习来学习攻击策略。实验结果表明，即使服务器部署了先进的防御机制，该攻击框架也能实现强大的性能。我们的工作揭示了如何通过多代理协调攻击联邦学习系统
We propose a model-based reinforcement learning framework against federated learning systems. Our method first approximates the distribution of the aggregated data through cooperative multi-agent coordination. It then learns an attack policy through multi-agent reinforcement learning. Experimental results demonstrate that the proposed attack framework achieves strong performance even if the server deploys advanced defense mechanisms. Our work sheds light on how to attack federated learning systems through multi-agent coordination