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

Abstract

Personal authentication security is an essential area of research in privacy and cybersecurity. For individual verification, fingerprint and facial recognition have proved particularly useful. However, such technologies have flaws such as fingerprint fabrication and external impediments. Different AI-based technologies have been proposed to overcome forging or impersonating authentication concerns. Electrocardiogram (ECG)-based user authentication has recently attracted considerable curiosity from researchers. The Electrocardiogram is among the most reliable advanced techniques for authentication since, unlike other biometrics, it confirms that the individual is real and alive. This study utilizes a user authentication system based on electrocardiography (ECG) signals using deep learning algorithms. The ECG data are collected from users to create a unique biometric profile for each individual. The proposed methodology utilizes Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) to analyze the ECG data. The CNNs are trained to extract features from the ECG data, while the LSTM networks are used to model the temporal dependencies in the data. The evaluation of the performance of the proposed system is conducted through experiments. It demonstrates that it effectively identifies users based on their ECG data, achieving high accuracy rates. The suggested techniques obtained an overall accuracy of 98.34% for CNN and 99.69% for LSTM using the Physikalisch–Technische Bundesanstalt (PTB) database. Overall, the proposed system offers a secure and convenient method for user authentication using ECG data and deep learning algorithms. The approach has the potential to provide a secure and convenient method for user authentication in various applications.

What Else Can AI See in a Digital ECG?

“…These include the need to provide better quality ECG recordings, the need to develop more advanced deep learning models that can extract significant features from ECG data, the need to adopt secure data transmission protocols such as encryption and SSL (Secure Socket Layers), which can help protect the transmitted data, and the need to implement multifactor authentication, which requires multiple forms of identification in order to access the system (apart from the ECG, the system should require passwords or security tokens, facial image, fingerprints, etc.). It would be extremely important to regularly update and test the ECG-based deep learning authentication system to ensure its effective operation and satisfactory security [22].…”

“…Valvular diseases [11] Left ventricular hypertrophy [12] Onset of atrial fibrillation [13,14] Risk of sudden cardiac death [15] Biological age of the patient [16] Concentration of sex hormones [17] Heart transplant rejection [18] Cardiovascular complications of liver transplantation [19] Effects of cardiac resynchronization therapy [20] Pulmonary embolism [21] Identity verification [22]…”

MeshID: Few-Shot Finger Gesture Based User Identification Using Orthogonal Signal Interference

“…Vision technology [ 39 , 40 ] can identify different users through physical activity characteristics captured from image frames using high-resolution cameras, but it may easily fail in the conditions of luminous changes and obstacles placed in line-of-sight (LoS) [ 4 ], in particular raising severe user privacy concerns. Bioelectrical technology [ 41 , 42 , 43 , 44 ] can utilize bioelectrical sensors, e.g., electrocardiogram (ECG), electromyogram (EMG) and electroencephalogram (EEG), to precisely extract unique biomedical information through body’s electrical activities. Ashraf et al [ 45 ] propose a fusion system that uses biometric features of the iris and foot.…”

Comparison of Transfer Learning Techniques to Classify Brain Tumours Using MRI Images

“…The study contends that a set of vital metabolites aids in the differentiation of paediatric brain tumours. In another work by the authors in [9], several models were used including VGG-16, ResNet-50, and EfficientNet-B0. During comparative analysis, it was observed that EfficientNet-B0 outperforms the other models, yielding an accuracy of 98.8%.…”