Wireless Federated Learning with Local Differential Privacy
Wireless Federated Learning with Local Differential Privacy
复制标题
DOI：
10.1109/isit44484.2020.9174426
复制
发表时间：
2020-02
期刊：
影响因子：
--
通讯作者：
Mohamed Seif;R. Tandon;Ming Li
中科院分区：
文献类型：
--
作者：
Mohamed Seif;R. Tandon;Ming Li
在本文中，我们研究了联邦学习（FL）在无线信道上的问题，建模高斯多址接入信道（MAC），受本地差分隐私（LDP）的约束。我们发现，无线信道的叠加性质提供了带宽高效的梯度聚合的双重好处，结合强大的LDP保证的用户。我们提出了一个私人无线梯度聚合方案，这表明，当聚合梯度从K个用户，隐私泄漏每个用户规模为$\mathcal{O}\left（{\frac{1}{{\sqrt K } \right）$相比，正交传输中的隐私泄漏规模为一个常数。我们还提出了建议的私人FL聚合算法的收敛速度的分析和研究之间的权衡无线资源，收敛性和隐私。
In this paper, we study the problem of federated learning (FL) over a wireless channel, modeled by a Gaussian multiple access channel (MAC), subject to local differential privacy (LDP) constraints. We show that the superposition nature of the wireless channel provides a dual benefit of bandwidth efficient gradient aggregation, in conjunction with strong LDP guarantees for the users. We propose a private wireless gradient aggregation scheme, which shows that when aggregating gradients from K users, the privacy leakage per user scales as $\mathcal{O}\left( {\frac{1}{{\sqrt K }}} \right)$ compared to orthogonal transmission in which the privacy leakage scales as a constant. We also present analysis for the convergence rate of the proposed private FL aggregation algorithm and study the tradeoffs between wireless resources, convergence, and privacy.