Pre-image Resistant Cancelable Biometrics Scheme Using Bidirectional Memory Model

Cancelable biometrics is a promising template protection scheme which relies on encoding the raw biometric data using non-invertible transformation function. Existing cancelable biometrics schemes ensure recoverability of compromised templates as well as users’ privacy. However, these schemes cannot resist pre-image attacks. In this article, a pre-image resistant cancelable biometrics scheme is proposed, where associative memory is utilized to encode the cancelable transformation parameters with the privilege of high recognition performance. Bidirectional memory model has been suggested to memorize each user’s associated key using his biometric data based on association connectors. These connector values can be safely saved in the storage along with the cancelable biometric template. The cancelable template is generated using XOR operation between the biometric data and the associated key. The simulated experiments conducted on CASIA-IrisV3-Interval dataset show that the presented resistant scheme does not affect the classification power of the raw biometric data significantly. Moreover, the resistance of the presented scheme against complete or approximate disclosure of the raw biometric template is achieved.