Ensemble Siamese Network (ESN) Using ECG Signals for Human Authentication in Smart Healthcare System.

Area of Science:

• Biomedical Engineering
• Computer Science
• Cybersecurity

Background:

• Digital healthcare systems enable remote patient monitoring and visits.
• Continuous authentication enhances security for sensitive health data access.
• Existing machine learning authentication models face challenges with user enrollment and imbalanced datasets.

Purpose of the Study:

• To propose a novel authentication method using electrocardiogram (ECG) signals for digital healthcare systems.
• To develop an Ensemble Siamese Network (ESN) capable of handling subtle variations in ECG signals for robust authentication.
• To address limitations of current machine learning authentication models, including user enrollment and dataset imbalance.

Main Methods:

• Utilized ECG signals, readily available in digital healthcare, for user authentication.
• Developed and implemented an Ensemble Siamese Network (ESN) architecture.
• Incorporated preprocessing for feature extraction to enhance model performance.
• Trained and evaluated the ESN model on the ECG-ID and PTB benchmark datasets.

Main Results:

• Achieved high accuracy rates of 93.6% on the ECG-ID dataset and 96.8% on the PTB dataset.
• Obtained low equal error rates of 1.76% and 1.69% on the respective datasets.
• Demonstrated the model's effectiveness in handling small changes in ECG signals.

Conclusions:

• The proposed ESN model provides a robust, simple, and data-available solution for continuous authentication in digital healthcare.
• ECG-based authentication using the ESN is a promising approach for enhancing security in smart healthcare and telehealth.
• The method overcomes challenges associated with user enrollment and dataset imbalances in machine learning-based authentication.