| 引用本文: | 乔少杰,林羽丰,韩楠,杨国平,李贺,袁冠,毛睿,元昌安,Louis Alberto GUTIERREZ.基于贡献度证明共识机制的去中心化联邦学习框架.软件学报,2023,34(3):1148-1167 |
| |
|
| |
|
|
| 本文已被：浏览 3305次 下载 6200次 |
码上扫一扫！ |
|
|
| 基于贡献度证明共识机制的去中心化联邦学习框架 |
|
乔少杰1, 林羽丰1, 韩楠2, 杨国平1, 李贺3, 袁冠4, 毛睿5, 元昌安6, Louis Alberto GUTIERREZ7
|
|
1.成都信息工程大学 软件工程学院, 四川 成都 610225;2.成都信息工程大学 管理学院, 四川 成都 610225;3.西安电子科技大学 计算机科学与技术学院, 陕西 西安 710071;4.矿山数字化教育部工程研究中心(中国矿业大学), 江苏 徐州 221116;5.深圳大学 计算机与软件学院, 广东 深圳 518060;6.广西人机交互与智能决策重点实验室(广西科学院), 广西 南宁 530100;7.Department of Computer Science, Rensselaer Polytechnic Institute, New York, USA
|
|
| 摘要: |
| 在大数据背景下，保证数据可信共享是数据联邦的基本要求.区块链技术代替传统的主从架构，可以提高联邦学习（federated learning，FL）的安全性.然而，现有工作中，模型参数验证与数据持久化所产生的巨大通信成本和存储消耗，已经成为数据联邦中亟待解决的问题.针对上述问题，设计了一种高效的去中心化联邦学习框架（efficient decentralized federated learning framework，EDFL），能够降低存储开销，并显著提升FL的学习效率.首先，提出了一种基于贡献度证明（proof-of-contribution）的共识机制，使得区块生成者的选举基于历史贡献度而不采用竞争机制，从而有效发避免了挖矿过程产生的区块生成延迟，并以异步方式缓解模型参数验证中的阻塞问题；其次，提出了一种角色自适应激励算法，因为该算法基于节点的工作强度和EDFL所分配的角色，所以能够激励合法节点更积极地进行模型训练，并有效地识别出恶意节点；再者，提出一种区块链分区存储策略，使得多重局部修复编码块（local reconstruction code）可被均匀地分布到网络的各个节点上，进而降低节点的本地存储代价，并实现了较高的数据恢复效率；最后，在真实的FEMNIST数据集上，对EDFL的学习效率、存储可扩展性和安全性进行了评估.实验结果表明，EDFL在以上3个方面均优于主流的基于区块链的FL框架. |
| 关键词: 数据联邦 区块链 大数据安全管理 共识机制 存储策略 |
| DOI：10.13328/j.cnki.jos.006784 |
| 分类号: |
| 基金项目:国家自然科学基金（61772091，61802035，61962006）；四川省科技计划（2021JDJQ0021，2022YFG0186，2021YZD0009，2021ZYD0033）；成都市技术创新研发项目（2021-YF05-00491-SN，2021-YF05-02414-GX，2021-YF05-02413-GX，2021-YF05-02420-GX，2021-YF05-02424-GX）；成都市重大科技创新项目（2021-YF08-00156-GX，2021-YF08-00159-GX）；成都市“揭榜挂帅”科技项目（2021-JB00-00025-GX） |
|
| Decentralized Federated Learning Framework Based on Proof-of-contribution Consensus Mechanism |
|
QIAO Shao-Jie1, LIN Yu-Feng1, HAN Nan2, YANG Guo-Ping1, LI He3, YUAN Guan4, MAO Rui5, YUAN Chang-An6, Louis Alberto GUTIERREZ7
|
|
1.School of Software Engineering, Chengdu University of Information Technology, Chengdu 610225, China;2.School of Management, Chengdu University of Information Technology, Chengdu 610225, China;3.School of Computer Science and Technology, Xidian University, Xi'an 710071, China;4.Engineering Research Center of Mine Digitization (China University of Mining and Technology), Ministry of Education, Xuzhou 221116, China;5.College of Computer Science and Software Engineering, Shenzhen University, Shenzhen 518060, China;6.Guangxi Key Laboratory of Human-machine Interaction and Intelligent Decision (Guangxi Academy of Sciences), Nanning 530100, China;7.Department of Computer Science, Rensselaer Polytechnic Institute, New York, USA
|
| Abstract: |
| In the background of big data, ensuring credible data sharing is the basic requirement of data federation. Using blockchain technology to replace the traditional client-server architecture can improve the security of federated learning (FL). However, the huge communication cost and storage consumption generated by model parameter validation and data persistence in existing works have become problems that need to be solved urgently in data federation. To tackle these problems, an efficient decentralized federated learning framework (EDFL) is proposed, which can reduce the system overhead and significantly improve the learning efficiency of FL. Firstly, a consensus mechanism based on proof-of-contribution (PoC) is proposed where the election of the block generation is based on historical contribution instead of using the competition mechanism, thus, it can avoid the latency of the block generation caused by the mining process, and asynchronously alleviate the congestion in the model parameter validation. Secondly, a role-adaptive incentive algorithm is presented. Because the proposed algorithm is based on the work intensity of participating nodes and the role assigned by EDFL, it can motivate legitimate nodes to actively conduct model training and effectively identify malicious nodes. Thirdly, blockchain partition storage strategy is proposed. The proposed strategy enables multiple local reconstruction code chunks to be evenly distributed to nodes in the network, which reduces the local storage consumption and achieves higher efficiency of data recovery. Lastly, the learning efficiency, storage scalability, and security of EDFL are evaluated in real FEMNIST dataset. Experimental results show that EDFL outperforms the state-of-the-art blockchain-based FL framework from the above three aspects. |
| Key words: data federation blockchain big data security management consensus mechanism storage strategy |
|
|
|
|
