Cybersecurity in Healthcare: AI and Cloud Adoption

Abstract

The healthcare sector faces increasing cybersecurity threats, which traditional, reactive security measures cannot effectively handle. These threats pose risks to patient safety, data privacy, and operational continuity. This research addresses this issue by developing a framework that integrates Artificial Intelligence (AI) and cloud platforms to enable proactive, real-time threat detection and response while adhering to ethical standards in data privacy and security.
The study employs an explanatory sequential mixed-methods research design. The quantitative phase consists of an experimental evaluation of four AI-based anomaly detection models (Isolation Forest, Autoencoder, LSTM Autoencoder, and Transformer Autoencoder) on the UNSW-NB15 benchmark cybersecurity dataset. The qualitative phase involves a survey of 25 senior-level cybersecurity and IT professionals to gather insights on the practical challenges, strategic considerations, and best practices for implementing AI technologies in healthcare cybersecurity.
The quantitative results revealed significant performance variations among the models, with a critical trade-off between precision and recall. The Autoencoder model achieved high precision (94.25%) but low recall (38.48%), highlighting the challenge of balancing false positives and false negatives. The qualitative results indicated that the primary barriers to AI adoption are organizational and resource-based rather than technological. Key challenges include cost constraints (88%), integration with legacy systems (84%), and a lack of skilled professionals (80%). Experts emphasized the importance of a strategic approach for AI implementation, including foundational security and a human-in-the-loop approach.
While advanced AI models, especially Transformers, hold significant potential for enhancing cybersecurity, their successful implementation requires a strategic, human-centric approach. The research's primary contribution is the Proactive, Adaptive, and Resilient (PAR) Cybersecurity Framework, a model that combines AI-driven detection with strategic principles to help healthcare organizations build cybersecurity programs that are both technologically advanced and aligned with the mission of patient safety, while ensuring ethical data privacy standards.