Study of the combined variant of determination of attacks using neural network technologies

The modern world is impossible to imagine without computer networks: both local and global; therefore, the issue of network security becoming increasingly topical. Currently, methods detecting attacks can be strengthened by using neural networks, which confirms relevance topic. aim study a comparative analysis quality parameters combined variant consisting different networks. As research used: network; multilayer perceptron; Kohonen's self-organizing map. software implementation Kohonen map carried out in Python with wide range standard tools, creation perceptron fuzzy - Neural Network Toolbox packages, Fuzzy Logic system MatLAB. On created networks separately on their researches definition are out. It was determined that error first kind 11%, 4%, 10% 0%, second 7%, 6%, 9% 6% network, perceptron, version, respectively, proves feasibility option.