Cancelable Biometric System for face Recognition Based on a Regularized Restoration Model

Abstract

Now, we use the biometric systems instead of passwords or tokens in authentication applications in several fields to improve the security level. The advantage of a biometric system is that the biometric cannot be lost, because it is a part of human body. This work aims to secure biometrics by distorting and saving them in a database to keep the original biometrics away from hackers. In this scenario, even if the biometrics are stolen or hacked up, we can reuse them again by changing the distorted versions. This paper presents a scheme for face biometric distorsion using a regularization approach. This scheme begins with adding noise to the original faces, and then applying regularized reconstruction on the noisy face images to obtain face images with magnified fixed noise patterns. These versions can be used as cancelable templates.