I.N. Lavrinenko, N.I. Chervyakov, A.A. Yevdokimov, A.N. Golovko

In modern cryptographic hardware-software means symmetric algorithms of enciphering which elements are substitutions and rearrangements are widely applied. Growth of refinement cryptoanalitic attacks shows, that the probability of breaking of such codes becomes big. The problem of breaking of reports arises because of breaking actually cryptographic algorithm, and is critical in a case when time of operation of devices of the communications with the built - in reports of safety is great. In the given work possible complexities and the threats arising during updating of such cryptographic algorithms, working on the removed server are considered. For good safety such reports constant change of the code is offered. It is offered to apply neural networks to simplicity of performance of substitutions and rearrangements, and as a key to use their weights. It is shown, that at training neural networks transfer of a training set on an open liaison channel reduces stability of the code, and use of dissymmetric algorithms for closing a key demands the big hardware expenses. Updating existing weights of neuron is offered to a network. Such approach is much the best from the point of view of safety the decision. It is offered to organize the safe mechanism of dynamic updating of the cryptographic algorithm realized with the help of specially designed neural networks

1. Введение в криптографию /Под ред. В.В. Яценко. - М.:МЦНМО. ЧеРо. 1998. 272 с.
2. P.Kotlars, Z.Kotulski, Artificial intelligence methods in the present-day cryptographic, Proceeling of plouq-05, Zakopane 2005 (in Polish)/
3.  Mosanya E., Teucher Ch., Restero Н.F., Galley P., Sanchez E. CryptoBooster: A Reconfigurable and Modular Cryptographic Coprocessor, in: Cryptographic Hardware and Embedded Systems: CHES-99, LNCS 1717, Springer Berlin 1999.
4. Kotlars P., Kotulski Z. Neural network us a programmable block cipher, in: Kazimeier Wielky Universiti, Bydgoszcz, piotrk@ukw.edu.ru; Institute of Fundamental Nechnoloqical Research, PAS and Institute of Telecommunications, WVT zkotush@ippt.gov.ru.
5. Kotlars P., Kotulski Z. Application of neural networks for implementation of cryptographic functions, in: Multimedia in Business and Education, ISBN-9182218-7-0.
6. Kotlars P., Kotulski Z. On application of neural networks for S-box design, in: P.S. Szczepaniak, J.Kacprzyk, A.Niewiadomski, ed.Advances in Web Intelligence, AWIC 2005, LNCS 3528. Р. 243-248. Berlin 2005.
7. Копарцова А.Г., Максимов А.В. Нейрокомпьютеры: учебное пособие для вузов. М.: Издательство МГТУ. 2002. 320 с.
8. Kanter I., Kinzel W., Kanter E. Secure exchange of information by synchronization of neural networks. Europhys. Lett. 2002. 57. Р. 141.