Transformer-Based Intrusion Detection for Securing Medical Applications in 5G IoMT Networks

Abstract

The research proposes a transformer-based anomaly detection system that enhances the 5G IoMT system security in that it guards healthcare applications against patient data leakages. WUSTL- HDRL-2024 dataset simulates different 5G network communications along with attack scenarios, while the proposed model makes use of self-attention to detect Distributed Denial of Service (DDoS) and others such as Man-in-the-Middle (MiTM) and Ransomware and Buffer Overflow. The single-layer transformer obtained an AUC-ROC score of 0.9132 and an F1-score of 0.5019 in the task of DDoS attack detection but showed a medium effect in both tasks of MiTM and Ransomware attack detection due to imbalance among labels. As a way of conducting real-time operations and usage in resource-limited environment, it is possible to conduct analysis of aggregated IP flows through a sliding time window framework. Though, the model shows itself a promising safety solution for IoMT systems while encountering difficulties associated with real world data integration and interpretability challenges with unbalanced datasets.