基于区块链的公平性联邦学习模型

doi:10.19678/j.issn.1000-3428.0064095

摘要/Abstract

摘要： 为解决典型联邦学习框架在训练样本数据分布不均衡情况下产生的聚合模型对各个客户端模型不公平的问题，结合区块链的去中心化、不可篡改性以及智能合约的特点，提出基于本地数据特征的公平性联邦学习模型，以实现数据分布差异的客户模型可信安全共享。多个客户端通过区块链上传本地参数以及信用值，利用区块链的共识机制选择信用值最高的区块进行模型聚合，在模型聚合过程中按照节点信用依次进行融合，并根据区块链记录工作节点的本地模型参数作为证据，完成整体模型参数的聚合任务，在此基础上通过广播下传当前聚合模型参数，模型利用区块链的共识机制可降低参数在传输过程中所面临的安全风险。在开源数据集上的实验结果表明，该模型相较FedAvg模型训练精度提高40%，不仅能够优化非独立同分布下的模型训练精度，同时可以防止中间参数传输信息泄露，保证了多个客户端的利益与安全隐私，从而实现具有隐私保护的公平性模型。

关键词: 联邦学习, 区块链, 非独立同分布, 公平性, 隐私保护

Abstract: This paper focuses on the unfair problem that the aggregation model generates by the non-independent and homogeneous distribution of training data in a typical Federated Learning (FL) framework.Accordingly, the characteristics of decentralization, immutability, and smart contracts of blockchain are combined to propose a blockchain-based fairness FL model to achieve trusted and secure sharing of customer models with differences in data distribution.First, each client uploads local parameters and trust values through the blockchain.Subsequently, the consensus mechanism of the blockchain was utilized to select the block with the highest trust value for model aggregation.The chain records the local model parameters of the working nodes as evidence to complete the aggregation task of the overall model parameters.Ultimately, the proposed algorithm downloads the current model parameters by broadcasting with low-security risks in the parameter transmission process.Experiments under the training of the open-source dataset reveal that the training accuracy of this model is improved by 40% compared with the FedAvg model.Accordingly, besides improving the training accuracy with the non-independent data distribution of FL, the FL algorithm also prevents information leakage in intermediate parameter transmission.Thus, it is the fairness model of privacy protection which ensures the interests, security, and privacy of each client are upheld.

Key words: Federated Learning(FL), blockchain, non-independent identically distribution, fairness, privacy protection

中图分类号:

TP391

陈乃月, 金一, 李浥东, 蔡露鑫, 魏圆梦. 基于区块链的公平性联邦学习模型[J]. 计算机工程, 2022, 48(6): 33-41.

CHEN Naiyue, JIN Yi, LI Yidong, CAI Luxin, WEI Yuanmeng. Federated Learning Model with Fairness Based on Blockchain[J]. Computer Engineering, 2022, 48(6): 33-41.

http://www.ecice06.com/CN/Y2022/V48/I6/33

图/表 7

20220618112806

20220618112810

20220618112815

20220618112818

20220618112823

20220618112827

20220618112831

参考文献

[1] LIANG Y, GUO Y G, GONG Y X, et al.An isolated data island benchmark suite for federated learning[EB/OL].[2022-02-01].https://arxiv.org/abs/2008.07257.
[2] YANG Q, LIU Y, CHEN T J, et al.Federated machine learning:concept and applications[J].ACM Transactions on Intelligent Systems and Technology, 2019, 10(2):12.
[3] LIM W Y B, LUONG N C, HOANG D T, et al.Federated learning in mobile edge networks:a comprehensive survey[J].IEEE Communications Surveys & Tutorials, 2020, 22(3):2031-2063.
[4] GOODFELLOW I J, SHLENS J, SZEGEDY C.Explaining and harnessing adversarial examples[EB/OL].[2022-02-01].https://arxivpreprintarxiv:1412.6572.
[5] XIAO H, BIGGIO B, BROWN G, et al.Is feature selection secure against training data poisoning?[C]//Proceedings of IEEE International Conference on Machine Learning.Washington D.C., USA:IEEE Press, 2015:1689-1698.
[6] SHOKRI R, STRONATI M, SONG C Z, et al.Membership inference attacks against machine learning models[C]//Proceedings of IEEE Symposium on Security and Privacy.Washington D.C., USA:IEEE Press, 2017:3-18.
[7] POKHREL S R, CHOI J.Federated learning with blockchain for autonomous vehicles:analysis and design challenges[J].IEEE Transactions on Communications, 2020, 68(8):4734-4746.
[8] ZHANG C, XIE Y, BAI H, et al.A survey on federated learning[J].Knowledge-Based Systems, 2021, 216:106775.
[9] MOTHUKURI V, PARIZI R M, POURIYEH S, et al.A survey on security and privacy of federated learning[J].Future Generation Computer Systems, 2021, 115:619-640.
[10] MCMAHAN B, MOORE E, RAMAGE D, et al.Communication-efficient learning of deep networks from decentralized data[C]//Proceedings of PMLR'17.Washington D.C., USA:IEEE Press, 2017:1273-1282.
[11] 张亮, 刘百祥, 张如意, 等.区块链技术综述[J].计算机工程, 2019, 45(5):1-12. ZHANG L, LIU B X, ZHANG R Y, et al.Overview of blockchain technology[J].Computer Engineering, 2019, 45(5):1-12.(in Chinese)
[12] HAKAK S, KHAN W Z, GILKAR G A, et al.Securing smart cities through blockchain technology:architecture, requirements, and challenges[J].IEEE Network, 2020, 34(1):8-14.
[13] FAROUK A, ALAHMADI A, GHOSE S, et al.Blockchain platform for industrial healthcare:vision and future opportunities[J].Computer Communications, 2020, 154:223-235.
[14] LU Y L, HUANG X H, DAI Y Y, et al.Blockchain and federated learning for privacy-preserved data sharing in industrial IoT[J].IEEE Transactions on Industrial Informatics, 2020, 16(6):4177-4186.
[15] QU Y Y, GAO L X, LUAN T H, et al.Decentralized privacy using blockchain-enabled federated learning in fog computing[J].IEEE Internet of Things Journal, 2020, 7(6):5171-5183.
[16] WENG J S, WENG J, ZHANG J L, et al.DeepChain:auditable and privacy-preserving deep learning with blockchain-based incentive[J].IEEE Transactions on Dependable and Secure Computing, 2021, 18(5):2438-2455.
[17] SHARMA P K, PARK J H, CHO K.Blockchain and federated learning-based distributed computing defence framework for sustainable society[J].Sustainable Cities and Society, 2020, 59:102220.
[18] ZHAO W J.Blockchain technology:development and prospects[J].National Science Review, 2019, 6(2):369-373.
[19] XIE C, KOYEJO S, GUPTA I.Asynchronous federated optimization[EB/OL].[2022-02-01].https://arxiv.org/abs/1903.03934.
[20] JIANG Y H, KONEČNÝ J, RUSH K, et al.Improving federated learning personalization via model agnostic meta learning[EB/OL].[2022-02-01].https://arxiv.org/abs/1909.12488.
[21] LI T, SAHU A K, ZAHEER M, et al.Federated optimization in heterogeneous networks[C]//Proceedings of IEEE Symposium on Machine Learning and Systems.Washington D.C., USA:IEEE Press, 2020:429-450.
[22] ZHANG M, SAPRA K, FIDLER S, et al.Personalized federated learning with first order model optimization[EB/OL].[2022-02-01].https://arxiv.org/abs/2012.08565.
[23] GAO S, YU T Y, ZHU J M, et al.T-PBFT:an EigenTrust-based practical Byzantine fault tolerance consensus algorithm[J].China Communications, 2019, 16(12):111-123.
[24] 孙嘉豪, 孟翔斯, 张浩运, 等.基于改进PBFT的区块链知识产权保护模型[J].计算机工程, 2020, 46(12):134-141. SUN J H, MENG X S, ZHANG H Y, et al.Intellectual property protection model using blockchain based on improved PBFT[J].Computer Engineering, 2020, 46(12):134-141.(in Chinese)
[25] KIM H, PARK J, BENNIS M, et al.Blockchained on-device federated learning[J].IEEE Communications Letters, 2020, 24(6):1279-1283.
[26] 谭敏生, 杨杰, 丁琳, 等.区块链共识机制综述[J].计算机工程, 2020, 46(12):1-11. TAN M S, YANG J, DING L, et al.Review of consensus mechanism of blockchain[J].Computer Engineering, 2020, 46(12):1-11.(in Chinese)
[27] 王兵, 李辉灵, 牛新征.基于综合选举的DPoS共识算法[J].计算机工程, 2022, 48(6):50-56. WANG B, LI H L, NIU X Z.DPoS consensus algorithm with comprehensive election[J].Computer Engineering, 2022, 48(6):50-56.(in Chinese)
[28] NGUYEN T, KIM K.A survey about consensus algorithms used in Blockchain[J].Journal of Information Processing Systems, 2018, 14(1):101-128.
[29] MIKAVICA B, KOSTIĆ-LJUBISAVLJEVIĆ A.Blockchain-based solutions for security, privacy, and trust management in vehicular networks:a survey[J].The Journal of Supercomputing, 2021, 77(9):9520-9575.

选择文件类型/文献管理软件名称

选择包含的内容