Machine Learning Approaches for Predictive Analysis of Cybersecurity Threats in Telehealth Systems: A Systematic Review

Abstract

Background: In the dynamic technological environment, telehealth platforms experience growing vulnerability risks that originate from increased connectivity and adoption. Intelligent threat detection methods, such as machine learning, promise rapid responses to manage complex data and device assets supporting life-critical care services prone to cybersecurity challenges.
Methods: Six databases, IEEE Xplore, Google Scholar, PubMed, Scopus, Embase, Web of Science, and CINAHL, were searched to retrieve studies for performance metrics comparisons. A systematic literature review identified 4220 studies, of which 18 were selected for machine learning cybersecurity approaches applied in telehealth environments. The methodology was strengthened through screening, risk-of-bias assessments, the CASP Qualitative Checklist (2019), and the Keele et al. (2007) accumulated list, with adherence to PRISMA guidelines.
Results: Among the reviewed studies, 38.9% focused on supervised learning techniques, unsupervised learning methods at 21.74%, deep learning, at 22% and reinforcement learning at 13.04%.
Conclusions: This study's findings supported upgrading to machine learning security implementations, immediate investments, and indispensable improvements for telehealth ecosystems to safeguard against increasing data breaches and service-disruption threats that endanger patient safety and care delivery services.