Ідентифікація трафіку мереж передачі даних у реальному часі

The article discusses modern approaches to real-time identification and classification of network traffic. The research emphasizes the urgency of the problem associated with the high dynamics of the network environment and the need for accurate methods for managing, protecting data, and detecting anomalies. An overview of existing traffic models is presented, including models based on distributions, time series, fractal models, and Markov chains. Classification methods such as port numbers, DPI (Deep Packet Inspection), machine learning, and statistical packet analysis are analyzed. The critical analysis revealed the advantages and limitations of each model. The ease of implementation and effectiveness of distribution models are limited by the consideration of dependencies between data. Time series models are suitable for forecasting but require significant computing resources. Fractal models provide an analysis of long-term dependencies, but their application in real time is difficult. Markov chains demonstrate high accuracy but require accurate initialization of parameters. Particular attention is paid to the integration of machine learning and computing process optimization methods for working with encrypted traffic and in conditions of limited resources. The authors emphasize the need to develop hybrid solutions that combine the advantages of different approaches and identify prospects for further research aimed at creating more accurate and efficient traffic analysis systems. Key areas for further research include automating model parameterization, taking into account encrypted traffic, and developing machine learning algorithms optimized for real-time operation. This article is a significant contribution to the development of network traffic identification methods, suggesting ways to improve the security and quality of service of networks