A Data-Driven Ensemble Deep Learning Framework for Intelligent Intrusion Detection in Ethical Telehealth Systems

Abstract

The growing reliance on telehealth solutions has reshaped healthcare delivery while introducing new and persistent cybersecurity vulnerabilities. Telehealth systems operate across diverse networked medical devices and rely on continuous data exchange involving highly sensitive patient information, making them susceptible to increasingly complex cyber intrusions. Existing intrusion detection approaches often struggle to adapt to the dynamic and resource-constrained nature of these environments. This work presents a data-driven ensemble deep learning framework designed to enhance intrusion detection within ethically governed telehealth infrastructures. The proposed framework integrates Long Short-Term Memory networks, Convolutional Neural Networks, and Deep Belief Networks to capture sequential network behavior, extract discriminative traffic features, and model abstract attack patterns. Model training and evaluation are conducted using the UNSW-NB15 dataset, which represents contemporary attack vectors relevant to healthcare networks. The experimental findings demonstrate that the ensemble approach attains a detection accuracy of 79% while achieving a lower false positive rate than individual deep learning models, thereby improving operational reliability in clinical settings. Beyond performance outcomes, the framework emphasizes responsible artificial intelligence practices by supporting privacy protection, balanced classification decisions, and dependable system behavior. The results indicate that ensemble-based deep learning can provide an effective, trustworthy, and deployable intrusion detection mechanism capable of strengthening cybersecurity resilience in modern telehealth systems.