Nowadays, the organization security against cyber-attacks is a matter of great importance and a challenging area, as it affects them financially and functionally. Novel attacks are emerging daily, threatening a large number of businesses around the world. For this reason, the implementation and optimization of the performance of Intrusion Detection Systems is an urgent task. To solve this problem, the scientific community uses deep learning methods. In this paper, we pay special attention to attack detection methods built on different kinds of architectures, such as multilayer perceptron, gated recurrent unit, long short-term memory network, recurrent neural network, and convolutional neural network. To train and test their models, we used dataset UNSW-NB 15. The Australian Centre created this dataset for Cyber Security. It created to generate traffic, which is a hybrid of normal and attack activities. In finally we summarize this paper and discuss some ways to improve the performance of attack detection under thoughts of utilizing deep learning structures.Nowadays, the organization security against cyber-attacks is a matter of great importance and a challenging area, as it affects them financially and functionally. Novel attacks are emerging daily, threatening a large number of businesses around the world. For this reason, the implementation and optimization of the performance of Intrusion Detection Systems is an urgent task. To solve this problem, the scientific community uses deep learning methods. In this paper, we pay special attention to attack detection methods built on different kinds of architectures, such as multilayer perceptron, gated recurrent unit, long short-term memory network, recurrent neural network, and convolutional neural network. To train and test their models, we used dataset UNSW-NB 15. The Australian Centre created this dataset for Cyber Security. It created to generate traffic, which is a hybrid of normal and attack activities. In finally we summarize this paper and discuss some ways to improve the performance of attack detection under thoughts of utilizing deep learning structures.

Keywords: network traffic, computer attack, artificial neural network, traffic analysis, neural network configuration

One of the main conditions for ensuring information security is to prevent the spread of false and intentionally distorted information. Filtering the content of Internet information resources can serve as a solution to this problem. Recently, an approach using methods and mathematical models of artificial intelligence has been increasingly considered for the analysis and classification of disseminated data. The use of neural networks allows you to automate the process of processing a large array of information and connect a person only at the decision-making stage. The paper focuses on the learning process of a neural network. Various learning algorithms are considered: stochastic gradient descent, Adagrad, RMSProp, Adam, Adama and Nadam. The results of the implementation of text subject recognition using a recurrent neural network of the LSTM model are presented. The results of computational experiments are presented, an analysis is carried out and conclusions are drawn.

Keywords: information security, text analysis, artificial intelligence method, artificial neural network, recurrent LSTM network