[1]卫 琳,陈 超,韩 丁,等.公共数据授权运营的鲁棒联邦学习入侵检测系统[J].郑州大学学报(工学版),2027,48(XX):1-9.[doi:10.13705/j.issn.1671-6833.2027.01.008]
　WEI Lin,CHEN Chao,HAN Ding,et al.Federated Learning-based Robust Intrusion Detection System for Public Data Authorized Operations[J].Journal of Zhengzhou University (Engineering Science),2027,48(XX):1-9.[doi:10.13705/j.issn.1671-6833.2027.01.008]
点击复制

公共数据授权运营的鲁棒联邦学习入侵检测系统()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
48
期数:
2027年XX
页码:
1-9
栏目:
出版日期:
2027-12-10

文章信息/Info

Title:
Federated Learning-based Robust Intrusion Detection System for Public Data Authorized Operations
作者:
卫 琳1, 陈 超1, 韩 丁2, 李秋燕2, 和孟佯1
1.郑州大学 网络空间安全学院,河南郑州450001;2国网河南省电力公司经济技术研究院,河南郑州450052
Author(s):
WEI Lin1, CHEN Chao1, HAN Ding2, LI Qiuyan2, HE Mengyang1
1. School of Cyber Science and Engineering, Zhengzhou University, Zhengzhou 450001, China;2. Economic and Technical Research Institute of State Grid Henan Electric Power Company, Zhengzhou 4500 52, China
关键词:
聚类联邦学习; 公共数据; 入侵检测; 中毒攻击防御; 数据异构; 类别不平衡
Keywords:
clustered federated learning;  public data;  intrusion detection;  poisoning attack defense;  data heteroge; neity;  class imbalance
分类号:
TP18 TP391
DOI:
10.13705/j.issn.1671-6833.2027.01.008
文献标志码:
A
摘要:
公共数据授权运营在实现多方数据共享复用的同时,面临针对公共数据的恶意入侵威胁,亟须可靠的入侵检测技术保障数据安全与隐私。 传统基于联邦学习的入侵检测技术中存在以下问题:入侵检测数据类别不平衡,导致模型对少数类入侵样本误报率高;节点之间的数据异构性加剧了局部模型与全局模型的差异,导致收敛速度变慢;恶意节点可能通过中毒攻击破坏模型训练,影响良性节点的检测性能。 为此,提出了一种面向公共数据授权运营的鲁棒联邦学习入侵检测框架( FL-RIDS) 。 首先,引入了一种基于有效样本数加权的入侵检测损失函数(ESLoss) ,通过动态调整类别权重显著提升了模型对小样本入侵类别的识别敏感度;其次,提出一种基于模型性能评估的聚类方法( SEC) ,通过在云服务器上评估各主体节点本地模型表现,将数据特征相似的节点进行聚类,有效缓解了异构数据导致的模型性能下降和收敛问题;最后,为确保系统对抗恶意攻击的鲁棒性,提出了一种轻量级中毒检测算法( LPD) ,通过监测主体节点本地模型与集群内全局模型分类器更新方向的相似度变化,动态识别训练过程中发动中毒攻击的恶意节点。 实验结果表明:在 CICIDS2017 和 CAN-intrusion-dataset 两个数据集上,ESLoss 相对交叉熵损失函数 f1score 最高分别提升了 18. 9 和 3. 2 百分点;SEC 与 FedAvg 相对比,f1score 最高分别提升了 4. 9和 5. 9 百分点,并展示出良好的抗中毒攻击能力;同时,LPD 与 FedAvg 相对比,f1score 最高分别提升了 54. 2 和 68. 4百分点,能有效抵抗恶意节点在训练中途的中毒攻击。
Abstract:
While public data authorized operation enables multi-party data sharing and reuse, it confronts malicious intrusion threats targeting public data, making reliable intrusion detection technologies imperative to guarantee datasecurity and privacy. Traditional federated learning-based intrusion detection technologies have three major limitations: severe class imbalance of intrusion detection data leading to high false positive rates for minority intrusionsamples; data heterogeneity across nodes exacerbating deviations between local and global models and slowing down model convergence; and poisoning attacks from malicious nodes undermining model training and degrading the detection performance of benign nodes. A federated learning-based robust intrusion detection system ( FL-RIDS) for public data authorized operation was proposed. Firstly, an effective sample number weighted loss function ( ESLoss) for intrusion detection was designed to dynamically adjust class weights and significantly improve the model′s sensitivity to few-sample intrusion categories. Secondly, a model performance evaluation-based clustering method( SEC) was proposed. It evaluated the performance of local models of all participant nodes on the cloud server and clusters nodes with similar data features, effectively mitigating performance degradation and slow convergence caused by data heterogeneity. Finally, a lightweight poisoning detection algorithm ( LPD) was proposed to enhance system robustness against malicious attacks. It dynamically identified malicious nodes launching poisoning attacks during training by monitoring the similarity variation of update directions between local node models and cluster global model classifiers. Experimental results showed that in CICIDS2017 and CAN-intrusion-dataset, ESLoss improved the f1score by 18. 9 and 3. 2 percentage points respectively at the highest level compared to the cross-entropy loss function. The SEC method improved the f1score by 4. 9 and 5. 9 percentage points respectively at the highest level compared to FedAvg with excellent anti-poisoning performance. In addition, LPD improved the f1score by 54. 2 and 68. 4 percentage points respectively at the highest level compared to FedAvg, which could effectivelyresist poisoning attacks launched by malicious nodes in the middle of model training.

参考文献/References:

[1] 魏明军, 李凤, 刘亚志, 等. 基于改进WGAN-GP和ResNet的车联网入侵检测方法[J]. 郑州大学学报（工学版）, 2024, 45(4): 30-37.

[2] Ileberi, E., Sun, Y. Advancing Model Performance With ADASYN and Recurrent Feature Elimination and Cross-Validation in Machine Learning-Assisted Credit Card Fraud Detection: A Comparative Analysis[J]. IEEE Access, 2024, 12: 133315-133327.

[3] Han, H., Wang, W.Y., Mao, B.H. Borderline-SMOTE: A New Over-Sampling Method in Imbalanced Data Sets Learning[C]//HUANG D S, ZHANG X P, HUANG G B. Advances in Intelligent Computing. Berlin, Heidelberg: Springer, 2005: 878-887.

[4] Chawla, N.V., Bowyer, K.W., Hall, L.O., et al. SMOTE: Synthetic Minority Over-sampling Technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321-357.

[5] Dina, A.S., Siddique, A.B., Manivannan, D. A deep learning approach for intrusion detection in Internet of Things using focal loss function[J]. Internet of Things, 2023, 22: 100699.

[6] Sheikh, S.A., Sahidullah, M., Hirsch, F., et al. Advancing Stuttering Detection via Data Augmentation, Class-Balanced Loss and Multi-Contextual Deep Learning[J]. IEEE Journal of Biomedical and Health Informatics, 2023, 27(5): 2553-2564.

[7] McMahan, B., Moore, E., Ramage, D., et al. Communication-Efficient Learning of Deep Networks from Decentralized Data[C/OL]//Proceedings of the 20th International Conference on Artificial Intelligence and Statistics. PMLR, 2017: 1273-1282.

[8] Shan, Y., Yao, Y., Zhou, X., et al. CFL-IDS: An Effective Clustered Federated Learning Framework for Industrial Internet of Things Intrusion Detection[J]. IEEE Internet of Things Journal, 2024, 11(6): 10007-10019.

[9] Yi, L., Wang, G., Liu, X., et al. FedGH: Heterogeneous Federated Learning with Generalized Global Header[C/OL]//Proceedings of the 31st ACM International Conference on Multimedia. New York, NY, USA: Association for Computing Machinery, 2023: 8686-8696.

[10] Ghosh, A., Chung, J., Yin, D., et al. An Efficient Framework for Clustered Federated Learning[J]. IEEE Transactions on Information Theory, 2022, 68(12): 8076-8091.

[11] Han, W., Peng, J., Yu, J., et al. Heterogeneous Data-Aware Federated Learning for Intrusion Detection Systems via Meta-Sampling in Artificial Intelligence of Things[J]. IEEE Internet of Things Journal, 2024, 11(8): 13340-13354.

[12] Duan, M., Liu, D., Ji, X., et al. FedGroup: Efficient Federated Learning via Decomposed Similarity-Based Clustering[C/OL]//2021 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Big Data & Cloud Computing, Sustainable Computing & Communications, Social Computing & Networking (ISPA/BDCloud/SocialCom/SustainCom). 2021: 228-237.

[13] Li, H., Sun, X., Zheng, Z. Learning to Attack Federated Learning: A Model-based Reinforcement Learning Attack Framework[J]. Advances in Neural Information Processing Systems, 2022, 35: 35007-35020.

[14] Baruch, G., Baruch, M., Goldberg, Y. A Little Is Enough: Circumventing Defenses For Distributed Learning[C/OL]//Advances in Neural Information Processing Systems: Vol. 32. Curran Associates, Inc., 2019 .

[15] Cao, X., Fang, M., Liu, J., et al. FLTrust: Byzantine-robust Federated Learning via Trust Bootstrapping[EB/OL]. (2020-12-27).

[16] Yin, D., Chen, Y., Kannan, R., et al. Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates[C/OL]//Proceedings of the 35th International Conference on Machine Learning. PMLR, 2018: 5650-5659.

[17] Cui, Y., Jia, M., Lin, T.Y., et al. Class-Balanced Loss Based on Effective Number of Samples[C/OL]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019: 9260-9269.

[18] Rousseeuw, P.J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis[J]. Journal of Computational and Applied Mathematics, 1987, 20: 53-65.

[19] Seo, E., Song, H.M., Kim, H.K. GIDS: GAN based Intrusion Detection System for In-Vehicle Network[C/OL]//2018 16th Annual Conference on Privacy, Security and Trust (PST). 2018: 1-6.

[20] Rosay, A., Carlier, F., Leroux, P. MLP4NIDS: An Efficient MLP-Based Network Intrusion Detection for CICIDS2017 Dataset[C]//BOUMERDASSI S, RENAULT É, MÜHLETHALER P. Machine Learning for Networking. Cham: Springer International Publishing, 2020: 240-254.

[21] Tuysuzoglu, G., Dogan, Y., Ozturk Kiyak, E., et al. Joint Tomek Links (JTL): An Innovative Approach to Noise Reduction for Enhanced Classification Performance[J]. IEEE Access, 2025, 13: 123059-123082.

[22] Tan, J., Li, B., Lu, X., et al. The Equalization Losses: Gradient-Driven Training for Long-tailed Object Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(11): 13876-13892.

[23] Li, T., Sahu, A.K., Zaheer, M., et al. Federated Optimization in Heterogeneous Networks[J]. Proceedings of Machine Learning and Systems, 2020, 2: 429-450.

[24] Thein, T.T., Shiraishi, Y., Morii, M. Personalized federated learning-based intrusion detection system: Poisoning attack and defense[J]. Future Generation Computer Systems, 2024, 153: 182-192.

更新日期/Last Update: 2026-10-30