Electrocardiogram (ECG)-Based User Authentication Using Deep Learning Algorithms.

Area of Science:

• Cybersecurity and Privacy Engineering
• Biometric Authentication Systems
• Artificial Intelligence in Security

Background:

• Traditional biometrics like fingerprints and facial recognition face challenges such as spoofing and environmental interference.
• Emerging biometric technologies are crucial for enhancing personal authentication security and privacy.
• Electrocardiogram (ECG) signals offer a unique, real-time biometric modality for robust user verification.

Purpose of the Study:

• To develop and evaluate a novel user authentication system leveraging electrocardiogram (ECG) signals.
• To investigate the efficacy of deep learning algorithms, specifically Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, for ECG-based authentication.
• To establish a secure and convenient biometric authentication method that overcomes limitations of conventional techniques.

Main Methods:

• Collection of ECG data from users to generate distinct biometric profiles.
• Application of Convolutional Neural Networks (CNNs) for feature extraction from ECG signals.
• Utilization of Long Short-Term Memory (LSTM) networks to model temporal dependencies within ECG data for authentication.

Main Results:

• The proposed system demonstrated high accuracy in user identification based on ECG data.
• CNN models achieved an accuracy rate of 98.34%.
• LSTM models achieved a superior accuracy rate of 99.69% on the PTB database.

Conclusions:

• ECG-based authentication using deep learning presents a secure and convenient alternative for personal identification.
• The combination of CNNs and LSTMs offers a powerful approach for analyzing ECG signals for biometric authentication.
• This methodology holds significant potential for enhancing security in diverse application scenarios requiring reliable user verification.