ECG-Based Authentication System with Enhanced Security Using Modified CNN Classifier

Recently, the electrocardiogram (ECG) authentication system has attained greater attention due to its pattern-based security. The ECG method provides individual cardiac patterns to achieve security. The proposed system has an ECG authentication by using deep learning (DL) techniques. The proposed system has two stages, feature extraction and classification. Initially, the hybrid module of Long Short-Term Memory (LSTM) network and an autoencoder is used for feature extraction. This encoder supports extracting the most relevant features from the ECG signals with excellent precision. Then, the modified Convolutional Neural Network (CNN) classifier is performed using extracted features. A Cuckoo Search optimization is applied to fine-tune the network’s hyperparameters to enhance the accuracy of the system. For a validation of results, the PhysioNet dataset is used and the proposed work overcomes existing models in terms of classification accuracy of 96% which has a better performance compared with traditional models.