Application of federated learning technology in modern telecommunication systems: optimization of network resources and cybersecurity

Abstract

The article explores the prospects and methodology for implementing Federated Learning (FL) technology in modern next-generation telecommunication networks (5G/6G). The conceptual principles of FL as an innovative approach to distributed machine learning are considered, which allows training global artificial intelligence models directly on peripheral devices (Edge computing) without the need for centralized transmission of raw data to processing servers. This approach significantly reduces the bandwidth requirements of backbone communication channels and minimizes the risks of leakage of confidential user information.
The architectural features of FL integration into telecommunication systems are analyzed, inclu-ding the interaction of peripheral nodes, base stations, and the central model aggregator. Particular attention is paid to federated aggregation algorithms, in particular FedAvg and its modifications, which ensure the efficient combination of locally trained parameters into a global model. The role of FL in increasing the level of cybersecurity of modern telecommunication systems is investigated. It is established that the use of decentralized learning allows significantly reducing the attack surface, since there is no need to transfer personalized data to centralized repositories. At the same time, poten-tial threats associated with model attacks (model poisoning, inference attacks) were analyzed, which require additional protection mechanisms and validation of updates.
It is shown that the implementation of Federated Learning in combination with Mobile Edge Computing (MEC) and Software Defined Networking (SDN) technologies creates the basis for building intelligent, adaptive and self-organizing telecommunication networks of the next generation. The results obtained indicate the high prospects of FL as a key technology for implementing the concept of networks with built-in artificial intelligence (AI-native networks), capable of operating effectively in conditions of high load dynamics and limited resources.

Keywords: telecommunication systems, Federated Learning, federated learning, edge computing (Edge Computing), 5G/6G, traffic optimization, cybersecurity.