标题： 用户级隐私保护联邦学习：分析与性能优化 显示英文标题

标题： User-Level Privacy-Preserving Federated Learning: Analysis and Performance Optimization

摘要： 联邦学习（FL）作为一种协作机器学习框架，能够在将数据训练成有用模型的同时，保护移动终端（MTs）的私有数据。 然而，从信息论的角度来看，好奇的服务器仍有可能通过MTs上传的共享模型推断出私有信息。 为了解决这个问题，我们首先利用局部差分隐私（LDP）的概念，并通过在上传到服务器之前向共享模型中添加人工噪声，提出了一种用户级差分隐私（UDP）算法。 根据我们的分析，UDP框架可以通过调整人工噪声过程的方差，实现对第$i$个MT的$(\epsilon_{i}, \delta_{i})$-LDP隐私保护水平。 随后，我们推导了UDP算法的理论收敛上界。 它表明存在一个最优的通信轮数以达到最佳的学习性能。 更重要的是，我们提出了一种通信轮数折扣（CRD）方法。 与启发式搜索方法相比，所提出的CRD方法可以在搜索的计算复杂性和收敛性能之间实现更好的权衡。 大量实验表明，使用所提CRD方法的UDP算法可以有效提高给定隐私保护水平下的训练效率和模型质量。 显示英文摘要

摘要： Federated learning (FL), as a type of collaborative machine learning framework, is capable of preserving private data from mobile terminals (MTs) while training the data into useful models. Nevertheless, from a viewpoint of information theory, it is still possible for a curious server to infer private information from the shared models uploaded by MTs. To address this problem, we first make use of the concept of local differential privacy (LDP), and propose a user-level differential privacy (UDP) algorithm by adding artificial noise to the shared models before uploading them to servers. According to our analysis, the UDP framework can realize $(\epsilon_{i}, \delta_{i})$-LDP for the $i$-th MT with adjustable privacy protection levels by varying the variances of the artificial noise processes. We then derive a theoretical convergence upper-bound for the UDP algorithm. It reveals that there exists an optimal number of communication rounds to achieve the best learning performance. More importantly, we propose a communication rounds discounting (CRD) method. Compared with the heuristic search method, the proposed CRD method can achieve a much better trade-off between the computational complexity of searching and the convergence performance. Extensive experiments indicate that our UDP algorithm using the proposed CRD method can effectively improve both the training efficiency and model quality for the given privacy protection levels.