Compression of network traffic parameters for detecting cyber attacks based on deep learning

The approach to compress network traffic parameters in systems for detecting attacks on telecommunication networks using the neural network and nonlinear principal components analysis is proposed. It is recommended to implement the deep neural network consists of a perceptron with lots of layers and that overcomes limitations of classic multilayer perceptron with the help of deep architecture. Experimental studies have confirmed that proposed approach provides a reduction of the data dimensionality and improves the accuracy of results during the analysis of network traffic parameters.