FREQUENCY CRYPTANALYSIS OF AN ASYMMETRIC CRYPTOGRAPHIC SYSTEM BASED ON ARTIFICIAL NEURAL NETWORKS AND NOISE-RESISTANT INFORMATION CODING

The purpose of this article is to describe a statistical attack on an asymmetric cryptosystem based on artificial neural networks and noise-resistant information coding, as well as to assess the practical applicability of this system in modern conditions, taking into account the possibility of carrying out this type of attack. The text of the work provides a step-by-step implementation of the attack and calculates the system's resistance to this type of cryptanalysis. The methodology of the study consists of mathematical modeling of a bit source with given probability parameters, as well as an analysis of the statistical characteristics of the values generated by it. Based on the analysis of the output values of the simulated source, the authors of the study derive a formula for calculating the resistance of the considered asymmetric cryptosystem to the described attack. They also come to the conclusion that the considered cryptosystem in the form in which it currently exists is extremely ineffective in modern conditions, has no practical application and is of purely academic interest. However, in conclusion, the authors note that if it is possible to limit the intruder's ability to obtain an unlimited number of " plaintext" / " ciphertext" pairs, the frequency cryptanalysis capability considered in this paper will be inapplicable. This will make it possible to consider the cryptosystem again as applicable in practice, at least from the point of view of the frequency cryptanalysis described in this paper.