Trans Secure Health: Transformer-Driven Cyber Threat Intelligence and Vulnerability Analysis for Medical Systems

Abstract

The rapid digital transformation of healthcare systems, driven by the adoption of Electronic Health Records (EHRs), interconnected medical devices, and integrated software platforms, has significantly enhanced service delivery and operational efficiency. However, this increased interconnectivity has also broadened the attack surface, resulting in a surge of cyber threats and vulnerabilities. Factors such as legacy infrastructure, system misconfigurations, and unpatched software further exacerbate the risk, potentially exposing sensitive patient data and critical healthcare services to security breaches. In this context, effective cyber-threat detection and vulnerability analysis have become crucial for ensuring the resilience of healthcare infrastructures. However, the growing volume of unstructured security reports and incident logs renders manual analysis inefficient, time-consuming, and prone to errors, limiting the ability to identify evolving threat patterns. To address these challenges, this study presents a Natural Language Processing (NLP)-driven framework for automated cyber-threat and vulnerability analysis, leveraging machine learning and transformer-based semantic modeling. Additionally, Adaptive Synthetic Sampling (AdaSYN) is employed to mitigate class imbalance issues and enhance model robustness. The extracted features are evaluated using multiple classification algorithms, including Greedy Tree Classifier (GTC), Tao Tree Classifier (TTC), KNearest Neighbors (KNN), and Gaussian Naive Bayes (GNB), facilitating comprehensive comparative analysis. Among these, the optimized GTC model is selected as the final predictive model to classify threat types, assign severity scores, and recommend appropriate mitigation strategies. The proposed system is implemented as a secure web-based application that enables efficient data upload, automated analysis, and intuitive result visualization.