Applied artifical neural network as a one-way function with a secret

Tarasenko Sergey Sergeevich

This article discusses the mechanism of obtaining various output data from an artificial neural network of direct propagation with immutable input data by adding uniform noise when calculating the activation functions of some part of the neurons of the network. The mechanism proposed in this paper can be used to construct a cryptoprimitive having the properties of a one-way function with a secret, which in turn the queue allows you to build asymmetric cryptosystems based on it. The calculation of the cryptographic resistance of this function to the calculation of its prototype is also performed and graphs reflecting the weak correlations between input and output sequences are clearly demonstrated. The paper presents the arguments. On the basis of which it is possible to put forward a hypothesis about the applicability of this cryptoprimitive for constructing cryptographic algorithms based on it that are resistant to cryptanalysis using quantum computing. The results obtained in the course of this study may be of value when considering the mathematical aspects of asymmetric cryptography based on error correction in noise-resistant codes.

Keywords:noise-resistant coding, Hamming distance, artificial neural networks, one-way function with secret, asymmetric cryptography.