Innovations in cyber defense could significantly improve the security of 5G networks

A framework for building stronger security into 5G wireless communications has been created by a Ph.D. student working in the Artificial Intelligence and Data Center at the University of Portsmouth.

5G is a critical component of intelligent systems and services such as healthcare and financial services because of its high network capacity and ability to quickly transmit large amounts of information between devices.

However, the dynamic nature of 5G networks, the large amount of data shared, and the constantly changing types of information being transmitted make these networks highly vulnerable to cyber threats and increase the risk of attack.

Dr. Hadiseh Rezaei, a University of Portsmouth student with a background in computer networks and software engineering, looked into this question and conducted experimental research that led to the development of a framework that could lead to better protection of information shared between devices using 5G networks.

The study, published in Computer Networks, proposes a new framework named FedLLMGuard. This framework combines two technologies: language and a large language model that understands patterns. The other is federated learning. It’s a system that learns from a variety of sources without sharing your personal information with anyone.

Together, they create a single system that accurately and quickly detects anomalies in 5G networks and securely protects data privacy in real-time.

Co-author Rahim Taheri, senior lecturer in computer science at the School of Computing at the University of Portsmouth, explained: “The majority of 5G intrusion detection systems still rely heavily on the numerical features of TV data, which limits their ability to capture textual, logical and contextual nuances.”

“Large language models are similar to the building blocks of data reading; they are trained on vast amounts of data, allowing them to understand language and context, but are still underutilized in network security. Federated learning, on the other hand, is a way to train AI models without humans seeing private data, providing a way to unlock information that can be fed into new AI applications.”

Hadise added, “Traditional intrusion detection systems often rely on fixed rules and static machine learning models. These approaches struggle to cope with the ever-changing nature of 5G traffic and are not effective against new or advanced attacks. However, FedLLMGuard dynamically adapts and defends against new threats as they emerge.”

“Through our experiments, we demonstrate that federated learning combined with large-scale language models can improve 5G security quickly and accurately. Think of it like a super-smart security guard for the internet that never tires, learns from every new trick hackers try, and protects everyone’s personal information at the same time.”

To prove how robust and reliable FedLLMGuard is, researchers tested it against a variety of cybersecurity threats. The framework has successfully defended against contradictory manipulation attempts, large-scale cyberattacks, stealth attacks designed to bypass security systems undetected, and data poisoning attacks that attempt to subvert the AI ​​training process.

FedLLMGuard outperformed all models, achieving 98.64% accuracy in recognizing security threats in less than 0.02 seconds (0.0113 seconds). The results demonstrate that FedLLMGuard has the ability to quickly detect threats, effectively mitigate attacks, and maintain high accuracy while ensuring data privacy, making it a scalable and resource-efficient security solution for 5G networks.

Recognizing that artificial intelligence (AI) and data science are rapidly advancing in the fields of research and innovation, the University of Portsmouth officially launched the Portsmouth Center for AI and Data Science (PAIDS) in June 2025.

At the heart of the PAIDS Center is the development of computer patterns, methods and algorithms to create solutions to improve systems and service delivery in areas such as health and welfare, education, cybersecurity and digital marketing.