Neural network system for converting the user biometric characteristics into a cryptographic key

Authors: Chuykov A. V., Vulfin A. M., Vasiliev V. I.

Annotation: The goal of this work is to improve the algorithms of biometric authentication by converting biometric features into a cryptographic key based on neural network technologies. To achieve this goal, an algorithm for the neural network transformation of the original biometric features into a cryptographic «private» key has been developed. The structure of the biometric cryptographic authentication system has also been developed, this structure is based on the use of neural network algorithms, distinguished by the method of forming and compressing the selected features using a self-organizing twodimensional Kohonen map, a convolutional neural network and the probabilistic method of principal components. To unambiguously match pairs of «biometric image» – «cryptographic key», bidirectional hetero-associative memory and a multilayer perceptron are used, that reduces the probability of a false negatives error and allows to achieve optimal results in the process of cryptographic key generation. As a result, there is no feasible possibility of compromising confidential data in the authentication process.

Keywords: biometric authentication, ppca, som, mlp, cnn, bam, alexnet