A Machine Learning Approach to Intrusion Detection in Cloud-Based e-Health Systems

Abstract

The integration of digital technologies and web-based services has significantly transformed healthcare delivery in recent years. However, this digital evolution has also increased the vulnerability of e-health systems to a wide range of cyber-attacks, posing serious risks to the confidentiality and integrity of patients’ electronic health records. This study proposes a machine learning-based intrusion detection approach specifically designed to secure cloud-based e-health systems. The Object-Oriented Analysis and Design Methodology (OOADM) was employed for system design and implementation. Two machine learning algorithms—Random Forest (RF) and K-Nearest Neighbor (KNN)—were applied to classify and detect various intrusion types, including General Practitioner attacks, Consent attacks, and uncorrelated threats. Simulations were conducted using Python and a medical dataset sourced from the Kaggle repository, partitioned into 80% training and 20% testing subsets with fine-tuned hyperparameters with the desired error value set to 0.000005 tolerance levels. The results of the model show that the Random Forest algorithm achieved a detection accuracy of 92%, significantly outperforming KNN algorithm, which achieved an accuracy level of 42%. The receiver operating characteristic (ROC) graph of KNN and RF classifiers indicated that the RF ROC curve has a value of 0.972 which is closer to 1 or top-left corner of the graph and is perfect and performed better than KNN model with a ROC value of 0.8892. These findings demonstrate the effectiveness of the RF algorithm in accurately identifying intrusions within cloud-hosted e-health systems, underscoring its potential for enhancing cybersecurity and safeguarding sensitive health information.