Federated Traffic-Driven Resource Allocation in Disaggregated Networks
Saroj Kumar Panda, Tania Panayiotou, Sadananda Behera, Georgios Ellinas · 2025
This work investigates fair federated traffic prediction across heterogeneous datasets for collaborative resource allocation in disaggregated optical networks. In such networks, where multiple operators independently manage different segments, federated learning (FL) enables privacy-preserving. However, ensuring fairness in FL remains challenging due to dataset heterogeneity arising from differences in network segment architectures, diverse traffic management policies, and user behaviors. This work examines fair FL for network traffic prediction by analyzing the trade-offs between fairness, global model accuracy, and training time as the number of federated datasets increases.