.png){kind=icon}
.png){kind=icon}
Return
Graph Neural Networks for IoMT Intrusion Detection: Modeling Architectures, Learning Paradigms, and Deployment Challenges
.png){kind=icon}
PRE
AI
H
F
G
M
S
DOI：10.1109/jiot.2026.3704452.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Internet of Medical Things (IoMT) connects medical sensors, devices, gateways, and cloud services for remote patient monitoring and timely care. However, its widespread application expands the attack surface. Many threat detection methods, including machine learning (ML) and traditional deep learning (DL), treat traffic as independent records, ignoring the correlation and evolution characteristics of network attacks. Graph neural network (GNN) methods have attracted significant attention for IoMT attack detection due to their unique advantages in handling graph structures. This article reviews GNN methods for IoMT threat detection and provides a comprehensive overview of key IoMT architecture components. It introduces threat classification across perception, network, service, and application layers. Technically, it covers aggregation-based, attention-based, and spatio–temporal hybrid models. It explores advanced paradigms like self-supervised learning (SSL), federated learning (FL), and knowledge distillation. This article summarizes GNN applications for IoMT and discusses key limitations, including data imbalance, heterogeneous and dynamic data challenges, and deployment bottlenecks on constrained devices. It also highlights robustness to adversarial manipulation, clinical workflow interpretability amidst false positives, and medical service availability. Finally, it discusses future research directions, including lightweight and scalable GNN methods, dynamic heterogeneous graph modeling, privacy-preserving federated architectures, few-shot learning, explainable-AI (XAI)-based interpretable GNN methods, and GNN methods robust to adversarial attacks.
Keywords:
Cyber security
deep learning (DL)
graph neural networks (GNNs)
Internet of Medical Things (IoMT)
intrusion detection
threat detection
Journal
IF:
8.9
Papers:
1.4W
Citations:
7.8W
Organization
Cited Papers