ИМИТАЦИЯ ПОВЕДЕНИЯ КОМПЬЮТЕРНОЙ СИСТЕМЫ С ПОМОЩЬЮ ИСКУССТВЕННЫХ НЕЙРОННЫХ СЕТЕЙ

Abstract

The design and the implementation of the simulation model of a computer system (CS) using artificial neural networks (ANNs) are considered. The purpose of the implementation is to create the easy-to-learn and easy-to-implement simulation model that allows to simulate both normal and anomalous processes in computer systems. The developed simulation model is the software, which consists of various modules combined using principles of the client-server architecture, allowing to run the model in both centralized and distributed modes of operation. The model allows to simulate the behavior of the CS with various topologies: a star, a tree, and a combination of these topologies. The main elements of the model are implemented in the form of four modules that fulfill their specific roles: an agent that generates data; a passive network element that transmits data with possible delays and losses; an active network element that processes and transmits data arriving at it, and the core – the central element of the model that receives data and sends it to additional modules for analysis. The modularity provides a high potential for further modifications of the simulation model by adding new modules. Using the generative adversarial network-based data generation module in the model makes it possible to generate data required for modeling the behavior of the studied CS. Based on the calculation of the Euclidean distance between matrices of transition probabilities of initial and generated data, it is shown that processes generated using the developed simulation model have a similar behavior with real ones. The designed model can be used to study the work of a real CS, including the imitation of an anomalous behavior.