Cancelable biometric authentication system based on ECG

Biometrics are widely deployed in various security systems; however, they have drawbacks in the form of leakage or stealing, therefore numerous solutions have been proposed to secure biometric template such as cancelable biometric, which is one of the possible solutions for canceling and securing biometric template. However, this problem is still open and to the best of our knowledge, few previous studies have proposed a complete authentic system using the cancelable biometric techniques based on electrocardiogram (ECG). In this paper, we have applied two cancelable biometric techniques for developing a human authentication system based on ECG signals. The first one is an improved Bio-Hashing and the second one is matrix operation technique. The improved Bio-Hash technique solves the problem of accuracy loss, which is the main drawback of basic Bio-Hash technique. The protected feature vector (Bio-Hashed code) is generated from the inner product between the ECG features matrix and tokenize number matrix. While the matrix operation technique is applied on the ECG feature matrix to produce a transformed template which is irreversible to the original features of the ECG. In the authentication stage, Feed-Forward Neural Network (FFNN) is used to verify individuals. After applying the two cancelable techniques on three public available ECG databases, experimental results show that the proposed system performs better regarding authentication and outperforms state-of-the-art techniques considered.