International Journal of Drug Delivery Technology
Volume 16, Issue 4s

Robust and Secure IoT Architecture for Machine Learning-Based Drug Administration in Remote Healthcare

Dr. Suhas S1, Dr. Lokesh S2, Pradeep Kumar H S3, Dr Palak Chaudhry4, Anil Kumar Jakkani5, Dr. Durairaj M6

1Assistant Professor, Department of Computer Science and Engineering, SJCE, JSS Science and Technology University, Mysore, Karnataka, India

Orcid ID: 0000-0003-0965-2809
2Associate Professor, Department of Computer Science and Engineering, The National Institute of Engineering, Mysore, Karnataka, India
3Assistant Professor, Department of Information Science, The National Institute of Engineering, Mysore
4Assistant Professor, Department of Rasa Shastra and Bhaishajya Kalpana, Parul Institute of Ayurved and Research, Parul University, Vadodara, Gujarat
5The Brilliant Research Foundation, India
6Associate Professor, Department of BioMedical Engineering, Vel Tech Multi Tech Dr. Rangarajan Dr. Sakunthala Engineering College, Avadi, Chennai-600062, Tamil Nadu

ABSTRACT

Remote patient care has been revolutionized by a fast-growing Internet of Medical Things (IoMT), which allows creating closed-loop therapeutic systems that automatically deliver drugs in response to real-time physiological measurements. Nevertheless, these mechanisms are confronted by a serious dual problem of not only the control of the doses precisely and as individual as possible but also of maintaining life-sensitive infrastructure against computer attacks that are becoming more and more advanced. The present paper presents the concept of a new and multi-layered IoT design, named SecureDose, which balances clinical efficacy with sound cybersecurity. SecureDose replaces intelligence to the Edge with a lightweight Long Short-Term Memory (LSTM) network to provide adaptive drug delivery and an ensemble Machine Learning-based Intrusion Detection System (IDS) to reduce threats in real-time.

The validity of the suggested architecture against conventional approaches is proved with experimental validation in terms of a Hardware-in-the-Loop (HIL) testbed. The LSTM-based controller demonstrated a Mean Absolute Error (MAE) of only 1.1mg/dl, which makes it significantly more effective than traditional PID controllers (4.2 mg/dL) and less likely to cause life-threatening drug overshoot to less than 1%. At the same time, Edge-hosted IDS has identified network traffic as malicious (99.2 percent), including DoS and Man-in-the-Middle attacks, and false positive rate was minimal (0.05). More importantly, an end-to-end latency of 112ms was sustained by the system, which is significantly below the 200ms safety margin necessary in critical care setting, demonstrating that very high-security requirements do not have to come at the cost of automating drug delivery systems.

Keywords: Internet of Medical Things (IoMT), Edge Computing, Intrusion Detection System (IDS), Automated Drug Delivery

How to cite this article: Suhas S, Lokesh S, Pradeep Kumar HS, Chaudhry P, Jakkani AK, Durairaj M, Robust and Secure IoT Architecture for Machine Learning-Based Drug Administration in Remote Healthcare. Int J Drug Deliv Technol. 2026;16(4s): 665-675; DOI: 10.25258/ijddt.16.4s.78

Source of support: Nil

Conflict of interest: None