multilayer neural networks

The article considers an algorithm for encrypting / decrypting text messages using multilayer neural networks (MLNN). The algorithm involves three steps: training a neural network based on the training pairs formed from a basic set of characters found in the text; encryption of the message using the weight coefficients of the hidden layers; its decryption using the weight coefficients of the output layer. The conditions necessary for successful encryption / decryption with this algorithm are formed, its limitations are emphasized.