Federated learning in network traffic classification: Taxonomy of implementation, application, and impact on sixth-generation wireless networks

Objectives

Network traffic classification (NTC) is crucial for network management. However, the surge in Internet traffic poses scalability and privacy issues. Researchers are turning to federated learning (FL) to tackle these issues. Despite the importance of FL, there is lack of literature that comprehensively reviews its implementation and application within NTC and its impact on Sixth-Generation Wireless (6G) networks. Current surveys cover the technical aspects of NTC without fully addressing the integration of FL and its broader implications. This study aims to review the technical implementation of FL in NTC and its application to various network-related areas, including 6G.

Methods

This study presents a taxonomy for FL implementation in NTC, considering aspects such as learning and aggregation approaches, topology, and client operations. It examines the limitations of these elements and their effects on performance, efficiency, scalability, and their impact on 6G. This study outlines a taxonomy for FL applications, focusing on privacy preservation, scalable classification, and shared security intelligence.

Novelty

The proposed taxonomy provides insights into research landscape and highlights its limitations. The analysis of the impact of FL-based NTC on 6G provides insight into its integration and implementation challenges. This study discusses open issues and advocates for future research directions in FL for NTC, including 6G.

Findings

The study identifies areas needing improvement such as privacy, addressing security, single-point-of-failure, hardware limitations, delays and heterogeneity concerns. The findings of this paper show that an optimal implementation approach is essential to cater for heterogeneity and real-time requirements of the network environment.