Classification of Cybersecurity Vulnerability Severity Levels in Healthcare Systems Using the Random Forest Algorithm
DOI:
https://doi.org/10.68300/injophars.v2i2.119Keywords:
cybersecurity, health systems, vulnerability, Random Forest, classification, machine learningAbstract
Cybersecurity is an important aspect of healthcare systems because vulnerabilities in devices, applications, and information technology infrastructure can threaten the confidentiality, integrity, and availability of healthcare information and services. This study evaluates the Random Forest algorithm for classifying cybersecurity vulnerability severity into four classes: LOW, MEDIUM, HIGH, and CRITICAL. The dataset used was the Healthcare Cybersecurity Vulnerabilities Dataset, containing 1,515 rows and 11 attributes. Eighteen records without a Severity label were excluded, leaving 1,497 labeled records. Text features were processed using Term Frequency–Inverse Document Frequency (TF-IDF), while categorical features were processed using One-Hot Encoding. The data were divided into training and testing sets using a stratified 80:20 split. The Random Forest model used 300 decision trees and class weighting to address class imbalance. On the 300-record test set, the model achieved 71.00% accuracy, 62.80% balanced accuracy, a 66.02% macro F1-score, and a 70.47% weighted F1-score. The highest recall was obtained for the MEDIUM class at 81.94%, whereas recall for the CRITICAL class reached only 43.33%. These results indicate that the model can correctly classify a substantial proportion of the records, but its performance is uneven across classes. Further evaluation through cross-validation, hyperparameter optimization, and comparison with other algorithms is required before the model can be considered for supporting vulnerability prioritization in healthcare systems.