Hybrid ANN–XGBoost Framework for Intelligent Intrusion Detection in Internet of Medical Things Networks
Keywords:
Internet of Medical Things, Intrusion Detection, Artificial Neural Network, XGBoost, Machine Learning, Healthcare Cybersecurity, Hybrid LearningAbstract
The Internet of Medical Things (IoMT) connects medical devices, sensors, monitoring platforms, and healthcare information systems, enabling continuous data exchange and remote healthcare services. However, increased connectivity expands the cybersecurity attack surface of healthcare infrastructures and exposes sensitive information and services to potential threats. This study presents a hybrid Artificial Neural Network–Extreme Gradient Boosting (ANN–XGBoost) framework for intelligent intrusion detection in IoMT environments. The proposed approach combines the nonlinear learning capability of Artificial Neural Network (ANN) with the ensemble learning capability of XGBoost through a weighted probability averaging mechanism. An experimental machine-learning workflow involving data acquisition, preprocessing, feature engineering, model development, hybrid integration, and comparative evaluation was adopted, together with an Agile-oriented software-development process. Logistic Regression, Random Forest, ANN, and standalone XGBoost were evaluated alongside the proposed hybrid model. Based on the experimental results, Logistic Regression, Random Forest, ANN, and XGBoost achieved accuracies of 95.4%, 97.6%, 97.1%, and 97.8%, respectively, while the hybrid ANN–XGBoost model achieved an accuracy of 98.6%, precision of 0.983, recall of 0.989, and F1-score of 0.986. The hybrid model also achieved a ROC-AUC of 0.99 and an average prediction response time of 1.18 seconds. A web-based implementation using Python, Flask, machine-learning libraries, and SQLite3 was developed to support authentication, data upload, preprocessing, prediction, and result logging. The study demonstrates a practical framework for integrating neural and gradient-boosting models for intelligent intrusion detection in IoMT environments.