CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
NEURAL NETWORK TECHNOLOGY FOR DETECTING ANOMALOUS NETWORK TRAFFIC
 |
Read | Chastikova Vera A., Zherlitsyn Sergey A., Volya Yana I., Sotnikov Vladimir V. NEURAL NETWORK TECHNOLOGY FOR DETECTING ANOMALOUS NETWORK TRAFFIC // Caspian journal : management and high technologies. — 2020. — №1. — pp. 20-32. |
/20-32.jpg)
Chastikova Vera A. - Kuban State Technological University, chastikova_va@mail.ru
Zherlitsyn Sergey A. - Kuban State Technological University, kpytooooo@gmail.com
Volya Yana I. - Kuban State Technological University, volya_y@mail.ru
Sotnikov Vladimir V. - Kuban State Technological University, bubert9@mail.ru
Existing network traffic analysis methods are considered. The relevance of the problem is demonstrated. The efficiency of swarm intelligence algorithms as applied to the task of training neural networks is analyzed, the features of these algorithms are revealed. An object - oriented library for detecting network attacks using a multi - layer perceptron neural network architecture has been implemented. The advantages and disadvantages of the implemented solution are described. A method is proposed for eliminating the widespread lack of datasets related to the imbalance of training data. The technologies used are described: the architecture of the neural network, the learning algorithm, a method of reducing the dimension of the processed data. A neural network model based on the LSTM architecture and embedding networks is proposed. To train the developed system, the use of the Adam algorithm based on gradient descent is proposed. Based on the use of the above algorithms, models and technologies, a software package for detecting network attacks was implemented and then tested.
Key words: нейронная сеть, сетевая атака, многослойный перцептрон, роевой интеллект, LSTMсеть, эмбеддинговая сеть, Focal Loss, алгоритм Adam, neural network, network attack, perceptron, swarm intelligence, LSTM network, embedding network, Focal Loss, Adam algorithm