A Novel Zero-Trust Machine Learning Green Architecture for Healthcare IoT Cybersecurity: Review, Analysis, and Implementation

TL;DR

The paper tackles the critical security risks of deploying IoT in healthcare by proposing a zero-trust, ML-driven architecture that emphasizes proactive threat detection and zero-day resilience. It introduces a two-part system (SM and IDH) with an autoencoder-based detector trained on CICIoT2023, implemented on low-power edge hardware (Raspberry Pi 4B) to balance accuracy, latency, and cost. The approach achieves up to $93.6\%$ mean accuracy with favorable energy and emission metrics ($7.5$W, $4.7\text{ mgCO}_2$ vs $725.9\text{ mgCO}_2$ in a baseline) and outperforms several hardware-centric baselines while remaining competitive with software-focused methods. The work contributes a concrete, deployable NSM solution tailored for healthcare IoT and demonstrates practical benefits in cost, power, and environmental impact, paving the way for broader edge deployments and richer threat datasets.

Abstract

The integration of Internet of Things (IoT) devices in healthcare applications has revolutionized patient care, monitoring, and data management. The Global IoT in Healthcare Market value is $252.2 Billion in 2023. However, the rapid involvement of these devices brings information security concerns that pose critical threats to patient privacy and the integrity of healthcare data. This paper introduces a novel machine learning (ML) based architecture explicitly designed to address and mitigate security vulnerabilities in IoT devices within healthcare applications. By leveraging advanced convolution ML architecture, the proposed architecture aims to proactively monitor and detect potential threats, ensuring the confidentiality and integrity of sensitive healthcare information while minimizing the cost and increasing the portability specialized for healthcare and emergency environments. The experimental results underscore the accuracy of up to 93.6% for predicting various attacks based on the results demonstrate a zero-day detection accuracy simulated using the CICIoT2023 dataset and reduces the cost by a factor of x10. The significance of our approach is in fortifying the security posture of IoT devices and maintaining a robust implementation of trustful healthcare systems.

Figures (7)

• Figure 1: IoT Components Share in Healthcare.
• Figure 2: IoT in Healthcare Market, by Component (USD M), 2019-2028 [1].
• Figure 3: Healthcare IoT framework block diagram.
• Figure 4: A block diagram of the common IoT Healthcare architecture model nasiri.
• Figure 5: Network Architecture for an abstract Autoencoder ML model.