Improving identification of malicious traffic in health and medical application: a novel study of weighting random forest model
Ahmed N. O. Algburi, İsa Avcı, Sulaiman M. Karim · IET conference proceedings. · 2025
In the rapidly evolving landscape of healthcare, the emergence of the Internet of Medical Things (IoMT) has promised unparalleled advancements in patient care. However, this technological revolution is accompanied by a shadowy undercurrent of cybersecurity threats that imperil the sanctity of medical data and patient privacy. This paper examines the multifaceted challenges posed by cyber-attacks within the IoMT framework, emphasizing the pervasive risk of physical harm to patients due to device vulnerabilities. Furthermore, the issue of unbalanced data exacerbates these vulnerabilities, rendering existing models and algorithms inadequate in mitigating the threat landscape effectively. Through a comprehensive analysis of these challenges, this paper underscores the urgent need for collaborative efforts to fortify cybersecurity defences and safeguard patient welfare in an increasingly interconnected healthcare ecosystem. To address class imbalance, a weighting random forest model was introduced (WRF). This model proposed a weighting component to the traditional Random Forest (RF). It emphasizes minority classes during training, as determined by class distribution entropy, hence boosting performance on imbalanced data sets. We trained the model on CIC-ISD2017 dataset to evaluate the model capacity, the test set for the evaluation consisted of 500 network traffic captures relating to medical settings whale the WRF achieved 99.98 accuracy.