EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Dual-Stream Transformer Framework for Secure and Fraud-Resilient Healthcare Systems

Farhan Shakil, Md Nahian Suhaimee, Md Omar Faruq, Md Rifat Al Amin Khan, Zeyar Aung, M. F. Mridha

IEEE Open Journal of the Computer Society · 2026 · 7 · 1678-1689

DOI: 10.1109/ojcs.2026.3746427 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of the Computer Society, 2026. Verfasst von Farhan Shakil und 5 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Khalifa University, Abu Dhabi, UAE.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Healthcare systems face dual challenges of financial fraud in medical claims and cybersecurity incidents targeting electronic health records (EHRs). These threats not only increase operational costs but also undermine trust, security, and patient safety. Traditional machine learning and deep learning approaches have been applied to address these issues, but most existing models treat fraud detection and cybersecurity anomaly detection as separate tasks. We propose a Dual-Stream Transformer Framework (DSTF) as a dual-task architecture that learns from healthcare claims and electronic medical record (EMR) access logs while preserving separate task-specific predictions. The framework employs two parallel transformer streams—one for healthcare claims and another for EMR access logs—coupled with cross-attention and a fusion layer to support representation-level information exchange and shared learning. We evaluate the proposed model on two benchmark datasets: the Healthcare Provider Fraud Detection dataset and the EMR Cyber Incident Detection dataset (combined version). Experimental results show that DSTF achieves an accuracy of 93.9%, F1-score of 0.905, and ROC–AUC of 0.963 for fraud detection, while reaching 91.2% accuracy, F1-score of 0.875, and PR–AUC of 0.872 for cybersecurity anomaly detection. In the joint fusion setting, DSTF maintains strong performance with 90.5% accuracy and 0.941 ROC–AUC, outperforming classical machine learning models, deep learning baselines, and state-of-the-art transformer architectures. These findings show that integrated transformer-based architectures can improve joint fraud and anomaly detection on the evaluated public claims and EMR-log datasets; however, deployment in national or region-specific health systems requires local data governance, schema mapping, privacy-preserving access controls, and external validation.

Förderung

  • Khalifa University, Abu Dhabi, UAE

Zitieren

APA

Shakil, F., Suhaimee, M. N., Faruq, M. O., Khan, M. R. A. A., Aung, Z., & Mridha, M. F. (2026). Dual-Stream Transformer Framework for Secure and Fraud-Resilient Healthcare Systems. IEEE Open Journal of the Computer Society, 7, 1678-1689. https://doi.org/10.1109/ojcs.2026.3746427

Vancouver

Shakil F, Suhaimee MN, Faruq MO, Khan MRAA, Aung Z, Mridha MF. Dual-Stream Transformer Framework for Secure and Fraud-Resilient Healthcare Systems. IEEE Open Journal of the Computer Society. 2026;7:1678-1689. doi:10.1109/ojcs.2026.3746427

BibTeX
@article{Shakil2026,
  author = {Shakil, Farhan and Suhaimee, Md Nahian and Faruq, Md Omar and Khan, Md Rifat Al Amin and Aung, Zeyar and Mridha, M. F.},
  title = {Dual-Stream Transformer Framework for Secure and Fraud-Resilient Healthcare Systems},
  journal = {IEEE Open Journal of the Computer Society},
  year = {2026},
  volume = {7},
  pages = {1678--1689},
  doi = {10.1109/ojcs.2026.3746427}
}
Kostenlos starten →