EUVIMEDEUVIMED entdecken
Crossref

Modeling of Transformer-Based Autoencoder Architecture for Secure IoT-Based Intrusion Detection in Critical Healthcare Systems

Saranya Ramachandran, Murugan Ambigapathy

International Journal of Medical Toxicology and Forensic Medicine · 16.09.2026

DOI: 10.22037/ijmtfm.v16.53130 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Medical Toxicology and Forensic Medicine, erschienen am 16.09.2026. Verfasst von Saranya Ramachandran und 1 weiteren Autor:innen.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

The Internet of Things (IoT) is widely used in healthcare for patient monitoring and continuous data collection. In intensive care unit (ICU) environments, many connected medical devices are used to monitor patient conditions in real time. Existing intrusion detection approaches using machine learning (ML) and deep learning (DL) have shown good results, but many still struggle with high-dimensional, complex network traffic data. Therefore, a more effective intrusion detection approach is required for secure IoT-based healthcare systems. The model is designed to distinguish between normal and malicious traffic in healthcare systems using deep learning approaches.

Methods

In the proposed model, relevant features are selected using a statistical and information-theoretic approach based on mutual information, variance thresholding, and correlation analysis to remove redundant information. For classification, a transformer-based autoencoder learns meaningful representations from network traffic, and a multilayer perceptron classifies the representations into normal and malicious classes. To improve model performance, Bayesian Optimization is leveraged for hyperparameter tuning.

Results

The proposed TAE-SHID model was evaluated using a benchmark dataset. The model achieved an accuracy of 96.86%, showing good classification performance compared with existing methods.

Conclusion

The results show that the proposed TAE-SHID model is effective for classification and outperforms existing methods.

Schlagwörter

intrusion detectionInternet of Things securityInternet of Medical Thingstransformer-based autoencodernetwork traffic classificationmalicious traffic detectionintensive care unitfeature selectionmutual informationBayesian optimization

Zitieren

APA

Ramachandran, S., & Ambigapathy, M. (2026). Modeling of Transformer-Based Autoencoder Architecture for Secure IoT-Based Intrusion Detection in Critical Healthcare Systems. International Journal of Medical Toxicology and Forensic Medicine. https://doi.org/10.22037/ijmtfm.v16.53130

Vancouver

Ramachandran S, Ambigapathy M. Modeling of Transformer-Based Autoencoder Architecture for Secure IoT-Based Intrusion Detection in Critical Healthcare Systems. International Journal of Medical Toxicology and Forensic Medicine. 2026. doi:10.22037/ijmtfm.v16.53130

BibTeX
@article{Ramachandran2026,
  author = {Ramachandran, Saranya and Ambigapathy, Murugan},
  title = {Modeling of Transformer-Based Autoencoder Architecture for Secure IoT-Based Intrusion Detection in Critical Healthcare Systems},
  journal = {International Journal of Medical Toxicology and Forensic Medicine},
  year = {2026},
  doi = {10.22037/ijmtfm.v16.53130}
}
Kostenlos starten →