Abstract
In recent years, the protection of human biometrics has witnessed an exponential growth. Fingerprint recognition has been utilized for cell phone authentication, biometric passports, and airport security. To improve the fingerprint recognition process, different approaches have been proposed. To keep biometrics away from hacking attempts, non-invertible transformations or encryption algorithms have been proposed to provide cancelable biometric templates for biometric protection. This paper presents a scheme that depends on chaos-based image encryption with different chaotic maps. The chaotic maps are used instead of the simple random number generator to overcome the loss of randomness in the case of a large number of images. To preserve the authentication performance, we should convolve the training images with random kernels to build the encrypted biometric templates. We can obtain different templates from the same biometrics by varying the chaotic map used to generate the convolution kernels. A comparative study is introduced between the used chaotic maps to determine the one, which gives the best performance. The simulation experiments reveal that the enhanced quadratic map 3 achieves the lowest error probability of 3.861% in the cancelable fingerprint recognition system. The cancelable fingerprint recognition system based on this chaotic map achieves the largest probability of detection of 96.139%, with an Equal Error Rate (EER) of 0.593.