《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (3): 856-863.DOI: 10.11772/j.issn.1001-9081.2024030296
收稿日期:
2024-03-20
修回日期:
2024-04-12
接受日期:
2024-04-16
发布日期:
2024-04-24
出版日期:
2025-03-10
通讯作者:
李京
作者简介:
林海力(1999—),男,浙江温州人,硕士研究生,主要研究方向:联邦学习
基金资助:
Received:
2024-03-20
Revised:
2024-04-12
Accepted:
2024-04-16
Online:
2024-04-24
Published:
2025-03-10
Contact:
Jing LI
About author:
LIN Haili, born in 1999, M. S. candidate. His research interests include federated learning.
Supported by:
摘要:
在对隐私保护的需求不断增长的当今社会,联邦学习正受到广泛关注。然而,在联邦学习中,服务器难以监管客户端的行为,致使懒惰客户端的存在为联邦学习的性能与公平性带来了潜在威胁。针对如何高效又准确地辨别懒惰客户端的问题,提出设置基于后门的双任务工作证明方法FedBD(FedBackDoor)。在FedBD中,服务器为参与联邦学习的客户端额外指定更易检测的后门任务,客户端在训练原任务的基础上训练后门任务,而服务器通过后门任务的训练情况间接监管客户端的行为。实验结果表明,在MNIST、CIFAR10等数据集上,相较于经典联邦平均算法FedAvg和先进算法GTG-Shapley(Guided Truncation Gradient Shapley),FedBD有一定优势。在CIFAR10数据集上,在懒惰客户端占比设置为15%时,FedBD比FedAvg的准确率提升可达10个百分点以上,比GTG-Shapley的准确率提升约2个百分点。此外,FedBD的平均训练时间仅为GTG-Shapley的11.8%,在懒惰客户端占比10%时辨别懒惰客户端的准确率可超过99%。可见,FedBD较好地解决了懒惰客户端难以监管的问题。
中图分类号:
林海力, 李京. 基于工作证明的联邦学习懒惰客户端识别方法[J]. 计算机应用, 2025, 45(3): 856-863.
Haili LIN, Jing LI. Lazy client identification method in federated learning based on proof-of-work[J]. Journal of Computer Applications, 2025, 45(3): 856-863.
方法 | 不同l下的准确率 | |||
---|---|---|---|---|
l=10% | l=20% | l=30% | l=40% | |
FedAvg | 0.585 3 | 0.473 8 | 0.402 8 | 0.387 6 |
MCS | 0.588 9 | 0.560 9 | 0.419 4 | 0.406 5 |
FedPCA | 0.646 7 | 0.613 3 | 0.476 5 | 0.423 8 |
GTG-S | 0.651 1 | 0.641 2 | 0.612 1 | 0.618 0 |
FedBD | 0.649 8 | 0.650 4 | 0.627 8 | 0.608 7 |
表1 不同懒惰客户端比例下不同方法的准确率
Tab. 1 Accuracies of different methods under different lazy client proportions
方法 | 不同l下的准确率 | |||
---|---|---|---|---|
l=10% | l=20% | l=30% | l=40% | |
FedAvg | 0.585 3 | 0.473 8 | 0.402 8 | 0.387 6 |
MCS | 0.588 9 | 0.560 9 | 0.419 4 | 0.406 5 |
FedPCA | 0.646 7 | 0.613 3 | 0.476 5 | 0.423 8 |
GTG-S | 0.651 1 | 0.641 2 | 0.612 1 | 0.618 0 |
FedBD | 0.649 8 | 0.650 4 | 0.627 8 | 0.608 7 |
数据集 | 方法 | 不同M和a下的准确率 | |||
---|---|---|---|---|---|
M=50, a=0.5 | M=50, a=0.2 | M=20, a=0.5 | M=20, a=0.2 | ||
MNIST | FedAvg | 0.868 8 | 0.823 5 | 0.895 4 | 0.853 5 |
MCS | 0.876 5 | 0.800 4 | 0.834 6 | 0.810 0 | |
FedPCA | 0.901 2 | 0.840 6 | 0.912 7 | 0.895 3 | |
GTG-S | 0.918 7 | 0.863 3 | 0.938 8 | 0.907 1 | |
FedBD | 0.907 6 | 0.895 4 | 0.944 4 | 0.919 9 | |
CIFAR10 | FedAvg | 0.521 0 | 0.490 4 | 0.549 4 | 0.536 5 |
MCS | 0.572 4 | 0.530 3 | 0.592 8 | 0.537 1 | |
FedPCA | 0.619 4 | 0.593 2 | 0.650 6 | 0.621 9 | |
GTG-S | 0.617 7 | 0.600 9 | 0.682 5 | 0.674 3 | |
FedBD | 0.639 7 | 0.591 8 | 0.672 7 | 0.678 0 |
表2 不同环境设置下不同方法的准确率
Tab. 2 Accuracies of different methods under different environment settings
数据集 | 方法 | 不同M和a下的准确率 | |||
---|---|---|---|---|---|
M=50, a=0.5 | M=50, a=0.2 | M=20, a=0.5 | M=20, a=0.2 | ||
MNIST | FedAvg | 0.868 8 | 0.823 5 | 0.895 4 | 0.853 5 |
MCS | 0.876 5 | 0.800 4 | 0.834 6 | 0.810 0 | |
FedPCA | 0.901 2 | 0.840 6 | 0.912 7 | 0.895 3 | |
GTG-S | 0.918 7 | 0.863 3 | 0.938 8 | 0.907 1 | |
FedBD | 0.907 6 | 0.895 4 | 0.944 4 | 0.919 9 | |
CIFAR10 | FedAvg | 0.521 0 | 0.490 4 | 0.549 4 | 0.536 5 |
MCS | 0.572 4 | 0.530 3 | 0.592 8 | 0.537 1 | |
FedPCA | 0.619 4 | 0.593 2 | 0.650 6 | 0.621 9 | |
GTG-S | 0.617 7 | 0.600 9 | 0.682 5 | 0.674 3 | |
FedBD | 0.639 7 | 0.591 8 | 0.672 7 | 0.678 0 |
方法 | 辨识准确率 | 辨识召回率 | ||
---|---|---|---|---|
l=10% | l=30% | l=10% | l=30% | |
MCS | 0.962 5 | 0.496 7 | 0.902 5 | 0.540 0 |
FedPCA | 0.977 5 | 0.792 5 | 0.948 3 | 0.678 7 |
GTG-S | 0.990 0 | 0.975 0 | 0.975 0 | 0.950 0 |
FedBD | 0.992 5 | 0.972 5 | 0.982 5 | 0.966 7 |
表3 不同懒惰客户端比例下不同方法辨识率的对比
Tab. 3 Comparison of different methods in identification rates under different lazy client proportions
方法 | 辨识准确率 | 辨识召回率 | ||
---|---|---|---|---|
l=10% | l=30% | l=10% | l=30% | |
MCS | 0.962 5 | 0.496 7 | 0.902 5 | 0.540 0 |
FedPCA | 0.977 5 | 0.792 5 | 0.948 3 | 0.678 7 |
GTG-S | 0.990 0 | 0.975 0 | 0.975 0 | 0.950 0 |
FedBD | 0.992 5 | 0.972 5 | 0.982 5 | 0.966 7 |
方法 | 训练时间 | 方法 | 训练时间 |
---|---|---|---|
FedAvg | 1.017 0 | FedBD | 5.053 5 |
MCS | 1.032 2 | GTG-S | 42.660 4 |
FedPCA | 3.992 8 |
表4 不同方法的训练时间对比 (s)
Tab. 4 Comparison of training time among different methods
方法 | 训练时间 | 方法 | 训练时间 |
---|---|---|---|
FedAvg | 1.017 0 | FedBD | 5.053 5 |
MCS | 1.032 2 | GTG-S | 42.660 4 |
FedPCA | 3.992 8 |
策略 | 辨识准确率 | 辨识召回率 |
---|---|---|
FedBD | 0.976 7 | 0.942 5 |
FedBD* | 0.920 0 | 0.912 5 |
表5 FedBD在不同策略下的辨识率对比
Tab. 5 Comparison of identification rate of FedBD under different strategies
策略 | 辨识准确率 | 辨识召回率 |
---|---|---|
FedBD | 0.976 7 | 0.942 5 |
FedBD* | 0.920 0 | 0.912 5 |
1 | McMAHAN H B, MOORE E, RAMAGE D, et al. Communication-efficient learning of deep networks from decentralized data [C]// Proceedings of the 20th International Conference on Artificial Intelligence and Statistics. New York: JMLR.org, 2017: 1273-1282. |
2 | 周传鑫,孙奕,汪德刚,等. 联邦学习研究综述[J]. 网络与信息安全学报, 2021, 7(5): 77-92. |
ZHOU C X, SUN Y, WANG D G, et al. Survey of federated learning research [J]. Chinese Journal of Network and Information Security, 2021, 7(5): 77-92. | |
3 | QIU J, WU Q, DING G, et al. A survey of machine learning for big data processing [J]. EURASIP Journal on Advances in Signal Processing, 2016, 2016: No.67. |
4 | LIN J, DU M, LIU J. Free-riders in federated learning: attacks and defenses[EB/OL]. [2023-10-28]. . |
5 | WANG G, DANG C X, ZHOU Z. Measure contribution of participants in federated learning [C]// Proceedings of the 2019 IEEE International Conference on Big Data. Piscataway: IEEE, 2019: 2597-2604. |
6 | JIA R, DAO D, WANG B, et al. Towards efficient data valuation based on the shapley value [C]// Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics. New York: JMLR.org, 2019: 1167-1176. |
7 | LIU Z, CHEN Y, YU H, et al. GTG-Shapley: efficient and accurate participant contribution evaluation in federated learning[J]. ACM Transactions on Intelligent Systems and Technology, 2022, 13(4): No.60. |
8 | XU X, LYU L, MA X, et al. Gradient driven rewards to guarantee fairness in collaborative machine learning [C]// Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2021: 16104-16117. |
9 | LV H, ZHENG Z, LUO T, et al. Data-free evaluation of user contributions in federated learning [C]// Proceedings of the 19th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks. Piscataway: IEEE, 2021: 1-8. |
10 | GU T, DOLAN-GAVITT B, GARG S. BadNets: identifying vulnerabilities in the machine learning model supply chain [J]. IEEE Access, 2019, 7: 47230-47244. |
11 | LI Y, JIANG Y, LI Z, et al. Backdoor learning: a survey [J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(1): 5-22. |
12 | ADI Y, BAUM C, CISSE M, et al. Turning your weakness into a strength: watermarking deep neural networks by backdooring [C]// Proceedings of the 27th USENIX Security Symposium. Berkeley: USENIX, 2018: 1615-1631. |
13 | LI J, SHAO Y, DING M, et al. Blockchain Assisted Decentralized Federated Learning (BLADE-FL) with lazy clients[EB/OL]. [2022-12-23]. . |
14 | 王勇,李国良,李开宇. 联邦学习贡献评估综述[J]. 软件学报, 2023, 34(3): 1168-1192. |
WANG Y, LI G L, LI K Y. Survey on contribution evaluation for federated learning [J]. Journal of Software, 2023, 34(3): 1168-1192. | |
15 | YU H, LIU Z, LIU Y, et al. A sustainable incentive scheme for federated learning [J]. IEEE Intelligent Systems, 2020, 35(4): 58-69. |
16 | ZHANG J, LI C, ROBLES-KELLY A, et al. Hierarchically fair federated learning [EB/OL]. [2023-05-01]. . |
17 | QI Y, HOSSAIN M S, NIE J, et al. Privacy-preserving blockchain-based federated learning for traffic flow prediction [J]. Future Generation Computer Systems, 2021, 117: 328-337. |
18 | KANG J, XIONG Z, NIYATO D, et al. Reliable federated learning for mobile networks [J]. IEEE Wireless Communications, 2020, 27(2): 72-80. |
19 | LIU Y, PENG J, KANG J, et al. A secure federated learning framework for 5G networks [J]. IEEE Wireless Communications, 2020, 27(4): 24-31. |
20 | LYU L, YU J, NANDAKUMAR K, et al. Towards fair and privacy-preserving federated deep models [J]. IEEE Transactions on Parallel and Distributed Systems, 2020, 31(11): 2524-2541. |
21 | PANDEY S R, TRAN N H, BENNIS M, et al. A crowdsourcing framework for on-device federated learning [J]. IEEE Transactions on Wireless Communications, 2020, 19(5): 3241-3256. |
22 | ZHAO B, LIU X, CHEN W N. When crowdsensing meets federated learning: privacy-preserving mobile crowdsensing system[EB/OL]. [2023-10-20]. . |
23 | CHEN Y, YANG X, QIN X, et al. Dealing with label quality disparity in federated learning [M]// YANG Q, FAN L, YU H. Federated learning: privacy and incentive, LNCS 12500. Cham: Springer, 2020: 108-121. |
24 | SHYN S K, KIM D, KIM K. FedCCEA: a practical approach of client contribution evaluation for federated learning [EB/OL].[2023-06-04]. . |
25 | RICHARDSON A, FILOS-RATSIKAS A, FALTINGS B. Rewarding high-quality data via influence functions [EB/OL]. [2022-08-30]. . |
26 | SHAPLEY L S. A value for n-person games [M]// KUHN H W, TUCKER A W. Contributions to the theory of games, volume Ⅱ, Annals of Mathematics Studies. Princeton: Princeton University Press, 1953: 307-318. |
27 | VAN CAMPEN T, HAMERS H, HUSSLAGE B, et al. A new approximation method for the Shapley value applied to the WTC 9/11 terrorist attack [J]. Social Network Analysis and Mining, 2018, 8: No.3. |
28 | GHORBANI A, ZOU J. Data shapley: equitable valuation of data for machine learning [C]// Proceedings of the 36th International Conference on Machine Learning. New York: JMLR.org, 2019: 2242-2251. |
29 | WANG T, RAUSCH J, ZHANG C, et al. A principled approach to data valuation for federated learning [M]// YANG Q, FAN L, YU H. Federated learning: privacy and incentive, LNCS 12500. Cham: Springer, 2020: 153-167. |
30 | SONG T, TONG Y, WEI S. Profit allocation for federated learning[C]// Proceedings of the 2019 IEEE International Conference on Big Data. Piscataway: IEEE, 2019: 2577-2586. |
31 | FAN Z, FANG H, ZHOU Z, et al. Improving fairness for data valuation in horizontal federated learning [C]// Proceedings of the IEEE 38th International Conference on Data Engineering. Piscataway: IEEE, 2022: 2440-2453. |
32 | FAN Z, FANG H, WANG X, et al. Fair and efficient contribution valuation for vertical federated learning [EB/OL]. [2024-04-07].. |
33 | YANG Q, LIU Y, CHEN T, et al. Federated machine learning: concept and applications [J]. ACM Transactions on Intelligent Systems and Technology, 2019, 10(2): No.12. |
34 | LI Y, ZHU L, JIA X, et al. Defending against model stealing via verifying embedded external features [C]// Proceedings of the 36th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2022: 1464-1472. |
35 | SOMMER D M, SONG L, WAGH S, et al. Towards probabilistic verification of machine unlearning [EB/OL]. [2023-12-01].. |
36 | LI Y, ZHANG Z, BAI J, et al. Open-sourced dataset protection via backdoor watermarking [EB/OL]. [2023-11-19]. . |
37 | ZHOU D W, YE H J, ZHAN D C. Learning placeholders for open-set recognition [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 4399-4408. |
38 | SALEHI M, MIRZAEI H, HENDRYCKS D, et al. A unified survey on anomaly, novelty, open-set, and out-of-distribution detection: solutions and future challenges [EB/OL]. [2023-12-03]. . |
39 | WANG Z, YANG E, SHEN L, et al. A comprehensive survey of forgetting in deep learning beyond continual learning [EB/OL]. [2022-12-23]. . |
40 | WANG L, ZHANG M, JIA Z, et al. AFEC: active forgetting of negative transfer in continual learning [C]// Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2021: 22379-22391. |
41 | ZHOU H, VANI A, LAROCHELLE H, et al. Fortuitous forgetting in connectionist networks [EB/OL]. [2023-12-18].. |
42 | LI X, XIONG H, AN H, et al. RIFLE: backpropagation in depth for deep transfer learning through Re-Initializing the Fully-connected LayEr [C]// Proceedings of the 37th International Conference on Machine Learning. New York: JMLR.org, 2020: 6010-6019. |
43 | FRENCH R M. Catastrophic forgetting in connectionist networks[J]. Trends in Cognitive Sciences, 1999, 3(4): 128-135. |
44 | LeCUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition [J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324. |
45 | KRIZHEVSKY A. Learning multiple layers of features from tiny images[R/OL]. [2023-09-14]. . |
46 | HSU T M H, QI H, BROWN M. Measuring the effects of non-identical data distribution for federated visual classification [EB/OL]. [2023-09-13]. . |
[1] | 曾辉, 熊诗雨, 狄永正, 史红周. 基于剪枝的大模型联邦参数高效微调技术[J]. 《计算机应用》唯一官方网站, 2025, 45(3): 715-724. |
[2] | 任志强, 陈学斌. 基于历史模型更新的自适应防御机制FedAud[J]. 《计算机应用》唯一官方网站, 2025, 45(2): 490-496. |
[3] | 徐超, 张淑芬, 陈海田, 彭璐璐, 张帅华. 基于自适应差分隐私与客户选择优化的联邦学习方法[J]. 《计算机应用》唯一官方网站, 2025, 45(2): 482-489. |
[4] | 王心妍, 杜嘉程, 钟李红, 徐旺旺, 刘伯宇, 佘维. 融合电力数据的纵向联邦学习企业排污预测模型[J]. 《计算机应用》唯一官方网站, 2025, 45(2): 518-525. |
[5] | 陈海田, 陈学斌, 马锐奎, 张帅华. 面向遥感数据的基于本地差分隐私的联邦学习隐私保护方案[J]. 《计算机应用》唯一官方网站, 2025, 45(2): 506-517. |
[6] | 张淑芬, 张宏扬, 任志强, 陈学斌. 联邦学习的公平性综述[J]. 《计算机应用》唯一官方网站, 2025, 45(1): 1-14. |
[7] | 朱亮, 慕京哲, 左洪强, 谷晶中, 朱付保. 基于联邦图神经网络的位置隐私保护推荐方案[J]. 《计算机应用》唯一官方网站, 2025, 45(1): 136-143. |
[8] | 晏燕, 钱星颖, 闫鹏斌, 杨杰. 位置大数据的联邦学习统计预测与差分隐私保护方法[J]. 《计算机应用》唯一官方网站, 2025, 45(1): 127-135. |
[9] | 陈廷伟, 张嘉诚, 王俊陆. 面向联邦学习的随机验证区块链构建[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2770-2776. |
[10] | 沈哲远, 杨珂珂, 李京. 基于双流神经网络的个性化联邦学习方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2319-2325. |
[11] | 罗玮, 刘金全, 张铮. 融合秘密分享技术的双重纵向联邦学习框架[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1872-1879. |
[12] | 陈学斌, 任志强, 张宏扬. 联邦学习中的安全威胁与防御措施综述[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1663-1672. |
[13] | 余孙婕, 曾辉, 熊诗雨, 史红周. 基于生成式对抗网络的联邦学习激励机制[J]. 《计算机应用》唯一官方网站, 2024, 44(2): 344-352. |
[14] | 张祖篡, 陈学斌, 高瑞, 邹元怀. 基于标签分类的联邦学习客户端选择方法[J]. 《计算机应用》唯一官方网站, 2024, 44(12): 3759-3765. |
[15] | 陈学斌, 屈昌盛. 面向联邦学习的后门攻击与防御综述[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3459-3469. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||