Abstract
This article introduces a new cancellable multi-biometric system based on the combination of a selective encryption method and a deep-learning-based fusion technology. The biometric face image is treated with an automatic face segmentation algorithm (Viola-Jones), and the image of the selected eye is XORed with a PRNG (Pseudo Random Number Generator) matrix. The output array is used to create a primary biometric template. This process changes the histogram of the selected eye image. Arnold's Cat Map is used to superimpose the PRN pixels only on the pixels of the primary image. Arnold's cat map deformed eyes are encrypted using the Advanced Encryption Standard (AES) to encrypt the biometric data stored in the database. In addition, the AES master key is used for the same person in the identity verification process to verify the biometric identity. It is created from the fingers of the right hand, and the right eye is integrated into this process using deep learning technology. The deep learning fusion process can prevent attacks on the biometric system as a whole. In order to avoid damage to the eye or fingerprint images, the design considers the other eye and fingerprint images.