Random Scrambling and Running Addition of PCA coefficients for Cancelable biometrics

In terms of user authentication, biometric systems offer numerous advantages in terms of convenience. However, they also suffer from disadvantages in some applications, for example, in the area of privacy protection. To enhance security and privacy in biometrics, cancelable biometrics have been introduced. In this paper, we propose cancelable biometrics for face recognition using a PCA method approach. In the proposed method, our experiment compares the performance between the cases where PCA coefficient vectors, and the transformed coefficient vectors are respectively used for verification. As a result, the proposed method satisfies a good level of condition for cancelable biometrics. The proposed method not only resolves a weak point of biometric systems, but also the method is simple.