基于模型特征方向的分层个性化联邦学习框架

doi:10.3969/j.issn.1671-1122.2025.06.004

信息网络安全 ›› 2025, Vol. 25 ›› Issue (6): 889-897.doi: 10.3969/j.issn.1671-1122.2025.06.004

• 专题论文: 网络主动防御 • 上一篇下一篇

基于模型特征方向的分层个性化联邦学习框架

邓东上¹, 王伟业², 张卫东², 吴宣够²()

1.重庆大学计算机学院，重庆 401331
2.安徽工业大学计算机科学与技术学院，马鞍山 243032

收稿日期:2025-02-28 出版日期:2025-06-10 发布日期:2025-07-11
通讯作者: 吴宣够 wuxgou@ahut.edu.cn
作者简介:邓东上（1998—），男，湖南，博士研究生，主要研究方向为物联网、群智感知、联邦学习|王伟业（1999—），男，安徽，硕士研究生，主要研究方向为个性化联邦学习|张卫东（1998—），男，安徽，博士研究生，主要研究方向为物联网、联邦学习|吴宣够（1979—），男，安徽，教授，博士，主要研究方向为物联网、联邦学习。
基金资助:
国家自然科学基金(62172003);安徽高校协同创新项目(GXXT-2022-051)

Layered Personalized Federated Learning Guided by Model Feature Orientation

DENG Dongshang¹, WANG Weiye², ZHANG Weidong², WU Xuangou²()

1. College of Computer Science, Chongqing University, Chongqing 401331, China
2. School of Computer Science and Technology, Anhui University of Technology, Ma’anshan 243032, China

Received:2025-02-28 Online:2025-06-10 Published:2025-07-11

摘要/Abstract

摘要：

随着人工智能和工业物联网的快速发展，工业智能化不断加速。用户隐私担忧的加剧使得联邦学习成为一种有前景的解决方案。然而，工业物联网中数据异构性、资源限制及潜在攻击者问题突出。现有的个性化联邦学习虽然为客户端定制模型，但忽略了联邦训练中的聚合偏差。为解决该问题，文章提出基于模型特征方向的分层个性化联邦学习框架，设计高效聚合机制，在不增加通信开销的前提下为各客户端精确捕获信息。该框架结合自适应模型量化机制确定聚合权重，应用分层动态聚合机制定制最优全局模型。在Fashion-MNIST和CWRU数据集上的实验结果表明，该方法显著优于4种基线方法，展现出更高的性能和效率。

关键词: 工业物联网, 隐私担忧, 个性化联邦学习

Abstract:

With the rapid advancement of artificial intelligence and the industrial Internet of Things (IIoT), industrial intelligence accelerates. Federated learning (FL) has emerged as a promising solution due to growing privacy concerns. However, FL in IIoT faces challenges like data heterogeneity, resource constraints, and adversarial threats. Existing personalized FL methods customized models for individual clients but ignored aggregation biases during training. To address this issue, this paper proposed a layered personalized FL framework guided by model feature orientations. The framework introduced an efficient aggregation mechanism that captured client-specific information without extra communication costs. It combined an adaptive model quantization mechanism for aggregation weights and a layered dynamic strategy to tailor global models for each client. Experiments on Fashion-MNIST and CWRU datasets show that our approach outperforms baselines in both performance and efficiency.

Key words: industrial internet of things, privacy concerns, personalized federated learning

中图分类号:

TP309

邓东上, 王伟业, 张卫东, 吴宣够. 基于模型特征方向的分层个性化联邦学习框架[J]. 信息网络安全, 2025, 25(6): 889-897.

DENG Dongshang, WANG Weiye, ZHANG Weidong, WU Xuangou. Layered Personalized Federated Learning Guided by Model Feature Orientation[J]. Netinfo Security, 2025, 25(6): 889-897.

图/表 7

图1

图2

图3

表1

图4

图5

图6

参考文献 24

[1]	LI Tian, HU Shengyuan, BEIRAMI A, et al. Ditto: Fair and Robust Federated Learning Through Personalization[C]// PMLR. International Conference on Machine Learning. New York: PMLR, 2021: 6357-6368.
[2]	FALLAH A, MOKHTARI A, OZDAGLAR A. Personalized Federated Learning with Theoretical Guarantees: A Model-Agnostic Meta-Learning Approach[EB/OL]. (2020-12-06)[2024-11-10]. https://dl.acm.org/doi/pdf/10.5555/3495724.3496024.
[3]	DINH C, TRAN N, NGUYEN T D. Personalized Federated Learning with Moreau Envelopes[C]// NIPS. Advances in Neural Information Processing Systems. New York: NIPS, 2020: 21394-21405.
[4]	JIANG Meirui, LE Anjie, LI Xiaoxiao, et al. Heterogeneous Personalized Federated Learning by Local-Global Updates Mixing via Convergence Rate[EB/OL]. (2024-08-07)[2024-11-10]. https://dblp.org/rec/conf/iclr/JiangLL024.html.
[5]	HUANG Yutao, CHU Lingyang, ZHOU Zirui, et al. Personalized Cross-Silo Federated Learning on Non-IID Data[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(9): 7865-7873.
[6]	CHEN Jingxue, YAN Hang, LIU Zhiyuan, et al. When Federated Learning Meets Privacy-Preserving Computation[J]. ACM Computing Surveys, 2024, 56(12): 1-36.
[7]	HUANG Wenke, YE Mang, SHI Zekun, et al. Federated Learning for Generalization, Robustness, Fairness: A Survey and Benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(12): 9387-9406.
[8]	FU Yuchuan, TANG Xinlong, LI Changle, et al. A Secure Personalized Federated Learning Algorithm for Autonomous Driving[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 25(12): 20378-20389.
[9]	ZHENG Tianyue, LI Ang, CHEN Zhe, et al. AutoFed: Heterogeneity-Aware Federated Multimodal Learning for Robust Autonomous Driving[C]// ACM. Proceedings of the 29th Annual International Conference on Mobile Computing and Networking. New York: ACM, 2023: 1-15.
[10]	MA Jialiang, TIAN Chunlin, LI Li, et al. FedMG: A Federated Multi-Global Optimization Framework for Autonomous Driving Control[C]// IEEE. 2024 IEEE/ACM 32nd International Symposium on Quality of Service (IWQoS). New York: IEEE, 2024: 1-10.
[11]	JIANG Guoqian, ZHAO Kai, LIU Xiufeng, et al. A Federated Learning Framework for Cloud-Edge Collaborative Fault Diagnosis of Wind Turbines[J]. IEEE Internet of Things Journal, 2024, 11(13): 23170-23185.
[12]	DENG Dongshang, ZHAO Wei, WU Xuangou, et al. DecFFD: A Personalized Federated Learning Framework for Cross-Location Fault Diagnosis[J]. IEEE Transactions on Industrial Informatics, 2024, 20(5): 7082-7091.
[13]	ZHANG Yudong, WANG Xu, WANG Pengkun, et al. Modeling Spatio-Temporal Mobility across Data Silos via Personalized Federated Learning[J]. IEEE Transactions on Mobile Computing, 2024, 23(12): 15289-15306.
[14]	ZHANG Yu, LU Hua, LIU Ning, et al. Personalized Federated Learning for Cross-City Traffic Prediction[EB/OL]. [2024-11-10]. https://doi.org/10.24963/ijcai.2024/611.
[15]	LI Qinbin, HE Bingsheng, SONG D. Model-Contrastive Federated Learning[C]// IEEE. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE, 2021: 10708-10717.
[16]	DURMUS A E, ZHAO Yue, RAMON M, et al. Federated Learning Based on Dynamic Regularization[EB/OL]. (2021-11-09)[2024-11-10]. https://arxiv.org/abs/2111.04263.
[17]	HU Ming, ZHOU Peiheng, YUE Zhihao, et al. FedCross: Towards Accurate Federated Learning via Multi-Model Cross-Aggregation[C]// IEEE. 2024 IEEE 40th International Conference on Data Engineering (ICDE). New York: IEEE, 2024: 2137-2150.
[18]	QI Peihan, ZHOU Xiaoyu, DING Yuanlei, et al. FedBKD: Heterogenous Federated Learning via Bidirectional Knowledge Distillation for Modulation Classification in IoT-Edge System[J]. IEEE Journal of Selected Topics in Signal Processing, 2023, 17(1): 189-204.
[19]	WANG Pengju, LIU Bochao, GUO Weijia, et al. Towards Personalized Federated Learning via Comprehensive Knowledge Distillation[C]// IEEE. 2024 IEEE International Conference on Systems, Man, and Cybernetics (SMC). New York: IEEE, 2024: 3807-3812.
[20]	LUO Jun, WU Shandong. Adapt to Adaptation: Learning Personalization for Cross-Silo Federated Learning[EB/OL]. [2024-11-10]. https://doi.org/10.24963/ijcai.2022/301.
[21]	MCMAHAN H B, MOORE E, RAMAGE D, et al. Communication-Efficient Learning of Deep Networks from Decentralized Data[EB/OL]. (2023-01-26)[2024-11-10]. https://arxiv.org/abs/1602.05629v4.
[22]	DENG Dongshang, WU Xuangou, ZHANG Tao, et al. FedASA: A Personalized Federated Learning with Adaptive Model Aggregation for Heterogeneous Mobile Edge Computing[J]. IEEE Transactions on Mobile Computing, 2024, 23(12): 14787-14802.
[23]	LI Tian, SAHU A K, ZAHEER M, et al. Federated Optimization in Heterogeneous Networks[J]. Proceedings of Machine Learning and Systems, 2020, 2: 429-450.
[24]	ARIVAZHAGAN M G, AGGARWAL V, SINGH A K, et al. Federated Learning with Personalization Layers[EB/OL]. (2019-12-02)[2024-11-10]. https://doi.org/10.48550/arXiv.1912.00818.

方法	Fashion-MNIST		CWRU
方法	$s=3$	$s=5$	两个位置	3个位置
FedAvg	94.32%	88.10%	83.38%	70.22%
FedProx	94.16%	87.86%	81.13%	68.92%
Ditto	94.09%	87.84%	91.00%	84.17%
FedPer	95.47%	91.81%	93.00%	91.42%
本文方法	95.74%	93.82%	95.56%	94.78%

基于模型特征方向的分层个性化联邦学习框架

Layered Personalized Federated Learning Guided by Model Feature Orientation

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 24

相关文章 6

编辑推荐

Metrics

本文评价

[1]	张学旺, 陈思宇, 罗欣悦, 雷志滔, 谢昊飞. 面向云辅助工业物联网的高效可搜索属性基加密方案[J]. 信息网络安全, 2024, 24(9): 1352-1363.
[2]	张晓均, 张楠, 郝云溥, 王周阳, 薛婧婷. 工业物联网系统基于混沌映射三因素认证与密钥协商协议[J]. 信息网络安全, 2024, 24(7): 1015-1026.
[3]	郭瑞, 魏鑫, 陈丽. 工业物联网环境下可外包的策略隐藏属性基加密方案[J]. 信息网络安全, 2023, 23(3): 1-12.
[4]	刘忻, 杨浩睿, 郭振斌, 王家寅. 一种实现在线注册与权限分离的工业物联网身份认证协议[J]. 信息网络安全, 2021, 21(7): 1-9.
[5]	刘忻, 郭振斌, 宋宇宸. 一种基于SGX的工业物联网身份认证协议[J]. 信息网络安全, 2021, 21(6): 1-10.
[6]	陈亚楠, 梅倩, 熊虎, 徐维祥. 适用于工业物联网的无证书并行密钥隔离签名[J]. 信息网络安全, 2018, 18(10): 1-9.