Machine Learning Techniques for Intrusion Detection Systems in 5G and Beyond Networks: A Systematic Literature Review
TL;DR
The paper concludes that no single model is universally best; rather, technique selection should depend on the type of attack, the quality of validation, the amount of latency, the scalability, privacy, and explainability, to determine the most appropriate approach.
Abstract
- 5G and subsequent network systems are designed with a number of these features like super-fast data transfer, ultra-low latency, huge number of connections, cloud network, multi-access edge computing, and network function virtualization. But also, they are the primary cause of the proliferation of cyber-attacks and the evolution of intelligent intrusion detection system. The manuscript describes a comprehensive review of the literature concerning the application of machine learning to intrusions detection systems in 5G and beyond networks. The survey includes various kinds of models, datasets, categories of attack, validation methods, assessment metrics, deployment readiness, explainability and privacy considerations. Studies published between 2020 and 2026 were examined using PRISMA-guided selection and structured data extraction method. Findings from 21 reviewed studies show that empirical ML/DL evaluation studies accounted for 42.9%, review or survey syntheses accounted for 23.8%, federated or transfer learning studies accounted for 19.0%, and dataset/testbed studies accounted for 14.3%. Representative benchmark accuracies ranged from 64.1% for application-layer PFCP detection to 100.0% for binary 5G-NIDD classification. The paper concludes that no single model is universally best; rather, technique selection should depend on the type of attack, the quality of validation, the amount of latency, the scalability, privacy, and explainability, to determine the most appropriate approach.