Design and Implementation of a Secure Framework for Biometric Identification Based on Convolutional Neural Network Technique

The implementation of robust authentication and identification methods has emerged as a pressing imperative to ensure the preservation of device integrity and the protection of sensitive data. Biometrics, especially electrocardiogram (ECG) technology, is presented as a promising solution due to its individualized and difficult-to-counterfeit nature. ECG signals offer advantages such as liveness detection and ubiquity. This study delves into ECG biometric recognition, leveraging Deep Learning advancements and employing wearable devices with ECG electrodes to ensure user-friendly scenarios. Samples were taken for 65 people, including males and females of varying ages. Six motivational cases were considered for each person. Subsequently, a model was constructed utilizing Sequential layers of a virtual convolutional neural network (CNN) and the Batch normalization layer, optimizing the training process for increased speed. The main contribution is a novel fingertip ECG identification system integrating Deep Learning and a deep convolutional neural network (CNN), trained on a large-scale database. The solution excels in speed and effectiveness, not requiring manual feature extraction or intricate model computations, and performs well with limited training data, achieving a good validation accuracy of 99.97%.