Artificial intelligence in 5G: optimization of network management and cybersecurity systems

The deployment of fifth-generation (5G) technologies is significantly changing the way telecommunications networks are designed and managed. The growth in the amount of data and the number of connected devices is also increasing the requirements for adaptive resource management, low latency, and guaranteed security, and traditional management mechanisms are often not flexible enough. In view of these developments, the use of artificial intelligence (AI) methods that can provide a dynamic response to changing conditions in real time is becoming increasingly important. [1]. The analysis of modern algorithms, in particular, such as: Graph Neural Networks (GNN), Deep Reinforcement Learning (DRL), as well as recurrent architectures LSTM and GRU, which are used to predict loads and optimize spectrum allocation. This approach allows us not only to reduce the risk of overload but also to increase the efficiency of network management in general [3]. The article considers the use of generative adversarial networks (GANs) in the tasks of detecting DDoS attacks. Our experimental results confirm that GAN models demonstrate high accuracy rates of up to 96%, outperforming traditional algorithms such as decision trees or SVMs, which is consistent with the results of Maleh et al. (2021) In addition, the article analyzes the possibilities of reducing energy consumption in 5G networks by implementing Multi-Agent Q-learning. This approach provides distributed decision-making at the base station level, taking into account local changes in the network, which helps to save energy without compromising the quality of service (Luo, 2020; Sutton & Barto, 2018). The research results show the significant potential of AI-oriented approaches in building scalable, resilient, and cyber threat-resistant next-generation networks. The recorded reduction in latency by 42%, improvement in spectral efficiency by 21%, and energy savings of up to 28% confirm the practical significance of the proposed solutions. The study suggests that it is possible to form a new architectural concept where AI not only supports the functioning of the network but also becomes a mechanism for its self-learning and evolution. [2].