Federated Learning for Compliance-Preserving Cyberattack Detection in Same Institution Internet of Medical Things (IoMT) Network

The study proposes a novel federated learning (FL) framework specifically designed for compliance-preserving cyber attack detection within a single healthcare institution's IoMT network. This framework enables collaborative threat detection across various internal departments or units (acting as federated clients) without requiring the direct sharing of raw patient medical data. Our approach implements and compares four state-of-the-art federated learning algorithms: FedAvg, FedProx, FedNova, and SCAFFOLD, across five different neural network architectures, including standard deep neural networks (DNN). The experiment methodology addresses realistic IoMT deployment scenarios within a healthcare institution, simulating heterogeneous data distributions from patient beds equipped with a total of 36 IoMT devices. This repository support the findings of the study, contains: Dataset: icu-dataset/ (Attack.csv, patient/environment monitoring.csv) Analysis notebook: FL-privacy-icu-iot.ipynb