端-边-云协同的联邦学习框架设计与实现

大理大学学报 ›› 2026, Vol. 11 ›› Issue (6): 19-26.

端-边-云协同的联邦学习框架设计与实现

大理大学数学与计算机学院，云南大理　671003

收稿日期:2025-03-27 出版日期:2026-06-15 发布日期:2026-06-30
通讯作者: 李晓伟，副教授，博士，E-mail：lixiaowei_xidian@163.com。
作者简介:杨悦朗，硕士研究生，主要从事数据隐私保护及共享研究。
基金资助:
国家自然科学基金项目（62262001；31960119）；云南省兴滇英才支持计划青年人才项目（20220137）

Design and Implementation of a Federated Learning Framework with End-Edge-Cloud Collaboration

College of Mathematics and Computer Science, Dali University, Dali, Yunnan 671003, China

Received:2025-03-27 Online:2026-06-15 Published:2026-06-30

摘要/Abstract

摘要： 随着数据隐私保护需求的日益增长，联邦学习凭借其无需共享原始数据即可实现联合模型训练的特性，已成为保护数据隐私的重要方法。然而，联邦学习在实际应用中仍面临诸多挑战，如恶意客户端的投毒攻击、通信成本高以及中心服务器计算负载过重等问题。为此，提出了端-边-云协同的联邦学习框架，通过分层管理客户端、边缘服务器和云服务器，有效降低云服务器的通信负担，提高系统的扩展性和效率。同时，在该框架下设计了结合基于声誉评估的局部模型聚合与基于全同态加密的全局模型聚合的联邦学习算法，以有效应对数据投毒攻击。实验结果表明，端-边-云协同的联邦学习框架不仅增强了系统对恶意攻击的防御能力，还有效降低了通信开销并提升了系统效率。

关键词: 联邦学习, 声誉机制, 全同态加密, 数据投毒攻击

Abstract: With the increasing demand for data privacy protection, federated learning has emerged as a crucial method for safeguarding data privacy due to its ability to enable joint model training without sharing raw data. However, federated learning still faces numerous challenges in practical applications, such as poisoning attacks by malicious clients, high communication costs, and excessive computational load on central servers. To address these issues, a federated learning framework with end-edge-cloud collaboration is proposed. By hierarchically managing clients, edge servers, and cloud servers, this framework effectively reduces the communication burden on cloud servers while improving system scalability and efficiency. Additionally, a federated learning algorithm was designed
within this framework, combining reputation-based local model aggregation and fully homomorphic encryption-based global model ag⁃
gregation to effectively counteract data poisoning attacks. Experimental results demonstrate that the federated learning framework with end-edge-cloud collaboration not only enhances the system's defense against malicious attacks but also significantly reduces communication overhead and improves system efficiency.

Key words: federated learning, reputation mechanism, fully homomorphic encryption, data poisoning attack

中图分类号:

TP311

杨悦朗, 王辛晨, 杨毅宇, 杨邓奇, 李晓伟. 端-边-云协同的联邦学习框架设计与实现[J]. 大理大学学报, 2026, 11(6): 19-26.

Yang Yuelang, Wang Xinchen, Yang Yiyu, Yang Dengqi, Li Xiaowei. Design and Implementation of a Federated Learning Framework with End-Edge-Cloud Collaboration[J]. Journal of Dali University, 2026, 11(6): 19-26.

参考文献

〔1〕 YANG Q， LIU Y， CHEN T J， et al. Federated Machine Learning： Concept and Applications〔J〕. ACM Transac⁃
tions on Intelligent Systems and Technology （TIST），2019， 10（2）： 1-19.
〔2〕 LI T， SAHU A K， TALWALKAR A， et al. Federated Learning： Challenges， Methods， and Future Directions〔J〕. IEEE Signal Processing Magazine， 2020， 37（3）： 50-60.
〔3〕 MCMAHAN B， MOORE E， RAMAGE D， et al. Communication-Efficient Learning of Deep Networks from Decentralized Data〔C〕//Artificial Intelligence and Statistics.PMLR， 2017： 1273-1282.
〔4〕 BAI P， ZHAI X， WU L， et al. FATE： A Federated Learning System for Analysis of Encrypted Data〔C〕//Proceedings of the 2019 Conference on Neural Information Processing Systems（ NeurIPS）. 2019：4423-4432.
〔5〕 LI Q B， HE B S， SONG D. Model-Contrastive Federated Learning〔C〕//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. June 20-25，2021， Nashville， TN， USA. IEEE， 2021： 10708-10717.
〔6〕 MELIS L， SONG C， DE CRISTOFARO E， et al. Exploiting Unintended Feature Leakage in Collaborative Learn⁃
ing〔C〕//2019 IEEE Symposium on Security and Privacy(SP）. IEEE， 2019： 691-706.
〔7〕 TOLPEGIN V， TRUEX S， GURSOY M E， et al. Data Poisoning Attacks Against Federated Learning Systems〔C〕//
Computer Security-ESORICS 2020. Cham： Springer，2020： 480-501.
〔8〕 ALSULAIMAWI Z. Federated Learning with Anomaly Detection via Gradient and Reconstruction Analysis〔EB/OL〕.〔 2025-01-05〕.https：//arxiv.org/abs/2403.10000.
〔9〕 ZHAO J Q， ZHU H， WANG F W， et al. Efficient and Privacy-Preserving Federated Learning against Poisoning Adversaries〔J〕. IEEE Transactions on Services Computing， 2024， 17（5）： 2320-2333.
〔10〕 LI X Y， YANG X， ZHOU Z C， et al. Efficiently Achieving Privacy Preservation and Poisoning Attack Resistance in Federated Learning〔J〕. IEEE Transactions on Information Forensics and Security， 2024， 19： 4358-4373.
〔11〕 MU X T， CHENG K， SHEN Y L， et al. FedDMC： Efficient and Robust Federated Learning via Detecting Malicious Clients〔J〕. IEEE Transactions on Dependable and Secure Computing，2024， 21（6）： 5259-5274.
〔12〕 LI Y T， WANG G J， WANG T， et al. FL-HSDP： Client-Side Defense for Federated Learning via Hessian Optimization
and Shuffled Differential Privacy〔C〕//2023 IEEE International Conference on High Performance Computing & Com⁃
munications， Data Science & Systems， Smart City & Dependability in Sensor， Cloud & Big Data Systems & Applica⁃
tion（ HPCC/DSS/SmartCity/DependSys）. December 17-21，2023， Melbourne， Australia. IEEE， 2024： 696-703.
〔13〕 MA Z R， MA J F， MIAO Y B， et al. ShieldFL： Mitigating Model Poisoning Attacks in Privacy-Preserving Federated Learning〔J〕. IEEE Transactions on Information Forensics and Security， 2022， 17： 1639-1654.
〔14〕刘少杰，文斌，王泽旭. 基于联邦学习的多技术融合数据交易方法〔J〕. 计算机工程， 2024， 50（3）： 182-190.
〔15〕裴浪涛，陈学斌，任志强，等. 面向多源数据的个性化联邦学习框架〔J〕. 计算机工程与应用， 2024， 60（19）： 278-287.
〔16〕LI Y. Enhancing Federated Learning Robustness and Fairness in Non-IID Scenarios〔D〕.Sydney： The University of
Sydney， 2024.