Нейрокриптографическая защита информации

Abstract

Along with the rapid development of digital communication technologies, which allowed to transmit messages in different forms over the network, the need to protect transmitted data from access by third parties has increased. One of the main ways to protect data is encryption. The main reason for encryption is that users must be aware of encryption methods and keys, without which information is meaningless in the form of symbols. As the efficiency of cryptanalysis methods has increased with computational performance, there has been a need for more sophisticated approaches to encryption. In particular, the use of such a promising approach as neural networks for data encryption – neurocryptography. Due to the fact that the increased power of technological tools continues to grow, today’s neural networks have been used in practice. Any encryption algorithm is based on generating different variants of a distorted code that can be recognized or reconstructed by a neural network with specified characteristics, and includes the following stages: preliminary, performing preliminary data processing and formation of a training sample; formation of a neural network, including training; and the main one, performing encryption or decryption. The article deals with the issue of increasing the efficiency of data protection by means of neurocryptography. Improving efficiency is achieved by selecting a group of cryptographic primitives, the implementation of which in the form of a neural network is the most effective. Efficiency in this case means the ratio of data encryption speed to the time of formation of the neural network.