Data Improvement Model Based on ECG Biometric for User Authentication and Identification.

Area of Science:

• Biometrics and Security
• Signal Processing
• Machine Learning for Healthcare

Background:

• Wearable technologies collect sensitive personal data, raising security concerns and making devices potential attack vectors.
• Biometric authentication, particularly using electrocardiogram (ECG) signals, is increasingly vital for online application security.
• Existing ECG biometric systems often lack scalability for large populations, limiting real-world applicability.

Purpose of the Study:

• To propose and validate a data improvement model, DETECT, for enhancing ECG-based biometric identification performance.
• To address the challenge of user identification with a high number of target classes in biometric systems.
• To improve the accuracy and reliability of ECG biometrics for large-scale authentication scenarios.

Main Methods:

• Investigated data enhancement techniques including increasing data examples and outlier removal.
• Incorporated additional relevant features to improve biometric identification accuracy.
• Developed and applied the DETECT (Data improvement model for ECG biometric identification) model to a large dataset.

Main Results:

• The DETECT model significantly increased precision from 78% to 92% for 1500 subjects and from 90% to 95% for 100 subjects.
• Achieved a low False Rejection Rate of 0.064003 and an exceptionally low False Acceptance Rate of 0.000033.
• Demonstrated improved performance in biometric identification with a higher number of target classes.

Conclusions:

• The proposed DETECT model effectively enhances ECG biometric identification for large-scale applications.
• Data improvement strategies are crucial for developing robust and scalable biometric security systems.
• The findings support the use of ECG biometrics as a reliable security measure for online authentication.