Federated Learning Approach for Edge Application of Network Link Classification Based on Traffic Matrices
IEEE Access · 2026 · 14 · 45487-45499
DOI: 10.1109/access.2026.3675110 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Nicolas André Alves Da Silva und 2 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Coordination for the Improvement of Higher Education Personnel.
Abstract (Originalsprache)
Monitoring network links is an important task for providing network services, such as routing in Software-Defined Networks (SDNs). Since the utilization level of a network link is directly related to the likelihood of packet congestion problems, monitoring the utilization levels of network links guarantees the outcome of tasks that are dependent on minimum requirements of network Quality of Service (QoS). Instead of using sensors on the network links throughout the monitoring, the SDN controller can calculate the utilization levels of its network links using end-to-end demand information and its own knowledge of the packet routes. To optimize this calculation, this article presents the NetLinkClassifier, which is a Machine Learning (ML) model that classifies a network link using a traffic matrix that contains all the end-to-end traffic demands of the network in a given time interval. The traffic matrix for the model input is assembled in a Multi-access Edge Computing (MEC) layer of the SDN, using traffic demand data from each SDN node. Before the model training, a supervised data set is distributively stored in this MEC layer. Therewith, the Proposed Model is trained using the Federated Learning (FL) paradigm. To guarantee the security in the data storage and transmissions, a cryptography system is recommended and a provided study evaluates this influence in the classification performance. Other performance reviews compare the use of the NetLinkClassifier model varying the model parameters and the SDN configurations. Finally, the article shows how the NetLinkClassifier model can be applied to support a Virtualized Network Function (VNF) that performs SDN packet routing.
Förderung
- Coordination for the Improvement of Higher Education Personnel
Zitieren
APA
Silva, N. A. A. D., Campos, C. A. V., & Lucena, S. C. D. (2026). Federated Learning Approach for Edge Application of Network Link Classification Based on Traffic Matrices. IEEE Access, 14, 45487-45499. https://doi.org/10.1109/access.2026.3675110
Vancouver
Silva NAAD, Campos CAV, Lucena SCD. Federated Learning Approach for Edge Application of Network Link Classification Based on Traffic Matrices. IEEE Access. 2026;14:45487-45499. doi:10.1109/access.2026.3675110
BibTeX
@article{Silva2026,
author = {Silva, Nicolas André Alves Da and Campos, Carlos Alberto Vieira and Lucena, Sidney Cunha De},
title = {Federated Learning Approach for Edge Application of Network Link Classification Based on Traffic Matrices},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {45487--45499},
doi = {10.1109/access.2026.3675110}
}
EUVIMED