Work

Modeling of zero trust with deep learning driven cyberattack detection system for enhanced medical IoT network security

Year: 2026

Type: article

Abstract: Zero trust architecture has emerged as a critical security paradigm for protecting sensitive medical data in smart healthcare systems. In IoT healthcare monitoring, this framework allows secure real-time data collection from connected medical devices while protecting sensitive patient information. Traditional security approaches based on machine learning (ML) and deep learning (DL) are widely adopted for anomaly detection and threat identification in such environments. However, these conventional models often rely on static training data, require significant computational resources, and struggle to adapt to rapidly evolving cyber-attacks in dynamic IoT healthcare settings. Consequently, they have less effective in providing adaptive and reliable security for continuously changing IoT healthcare environments. This study presents a Deep Learning-Driven Cyberattack Detection System Based on Zero Trust Architecture (DLCD-ZT). The proposed model initially preprocesses the input data to ensure feature consistency and data quality. A conditional mutual information maximization is used to reduce dimensionality while preserving the most informative feature dependencies. Then, the graph transformer network is integrated to learn structural dependencies and network behavior among IoT healthcare devices, enabling robust detection of malicious network activity. Finally, the AdaBelief optimizer is applied to improve convergence stability and enhance model training performance. Comprehensive experimental outcomes demonstrate the superiority of the proposed approach with a maximum accuracy of 93.26%, precision of 93.20%, recall of 92.96%, and F1-score of 86.16%, under the IoT Healthcare Security Dataset, confirming its effectiveness in achieving robust cyberattack detection in IoT environments. An ablation study is performed to assess the contribution of individual components in the proposed model. The results confirm the effectiveness of the DLCD-ZT framework for robust cyberattack detection and enhanced security in IoT healthcare environment.

Source: Scientific Reports

DOI: 10.1038/s41598-026-66788-4