Integrating NWDAF and Federated Learning for Privacy-Preserving Intrusion Detection in B5G IoHT Networks

The combination of Federated Learning and the Network Data Analytics Function (NWDAF) enables new possibilities for smart, secure, and privacy-aware network analysis in Beyond 5G (B5G) environments. In this work, we present a framework that explores this analysis to support distributed intrusion detection in the Internet of Healthcare Things (IoHT) scenario. Experimental results show that the system achieves stable convergence, with accuracy levels around 90% and high recall for major attack classes. Additionally, server-side Differential Privacy mechanisms maintain a reasonable privacy-utility balance. These findings demonstrate the feasibility of performing practical intrusion detection while preserving confidentiality in B5G-enabled IoHT environments.