When Federated Learning Meets Blockchain: A New Distributed Learning Paradigm
When Federated Learning Meets Blockchain: A New Distributed Learning Paradigm
复制标题
DOI：
10.1109/mci.2022.3180932
复制
发表时间：
2022-08-01
影响因子：
9
通讯作者：
Poor, H. Vincent
中科院分区：
文献类型：
--
作者：
Ma, Chuan;Li, Jun;Poor, H. Vincent
受终端用户设备日益强大的计算能力以及对共享敏感原始数据的日益增长的隐私担忧的推动，出现了一种被称为联邦学习（FL）的分布式机器学习范式。通过在每个客户端本地训练模型并在中央服务器上聚合学习模型，FL能够避免直接共享数据，从而减少隐私泄露。然而，传统的FL框架严重依赖于单个中央服务器，并且如果这样的服务器恶意行为，则其可能失败。为了解决这一单点故障，在这项工作中，研究了一个区块链辅助的去中心化FL框架，它可以防止恶意客户端毒化学习过程，从而为客户端提供一个自我激励和可靠的学习环境。在这个框架中，模型聚合过程是完全分散的，FL培训和区块链挖掘的任务被集成到每个参与者中。隐私和资源分配的问题进一步调查，在拟议的框架，和一个关键的和独特的问题，在拟议的框架中披露。特别地，懒惰客户端可以简单地复制由其他客户端共享的模型以获得益处，而无需将其资源贡献给FL。为了解决这些问题，提供分析和实验结果以阐明可能的解决方案，即，添加噪声以实现局部差分隐私，以及使用伪噪声（PN）序列作为水印来检测懒惰客户端。
Motivated by the increasingly powerful computing capabilities of end-user equipment, and by the growing privacy concerns over sharing sensitive raw data, a distributed machine learning paradigm known as federated learning (FL) has emerged. By training models locally at each client and aggregating learning models at a central server, FL has the capability to avoid sharing data directly, thereby reducing privacy leakage. However, the conventional FL framework relies heavily on a single central server, and it may fail if such a server behaves maliciously. To address this single point of failure, in this work, a blockchain-assisted decentralized FL framework is investigated, which can prevent malicious clients from poisoning the learning process, and thus provides a self-motivated and reliable learning environment for clients. In this framework, the model aggregation process is fully decentralized and the tasks of training for FL and mining for blockchain are integrated into each participant. Privacy and resource-allocation issues are further investigated in the proposed framework, and a critical and unique issue inherent in the proposed framework is disclosed. In particular, a lazy client can simply duplicate models shared by other clients to reap benefits without contributing its resources to FL. To address these issues, analytical and experimental results are provided to shed light on possible solutions, i.e., adding noise to achieve local differential privacy and using pseudo-noise (PN) sequences as watermarks to detect lazy clients.