Comparative analysis of machine and deep learning algorithms in network traffic anomaly detection

Abstract

The paper presents the results of research on the use of machine learning algorithms and neural networks for detecting anomalies in corporate network traffic. A representative set of test data is used to study the effectiveness of classification algorithms and their ensembles, as well as algorithms for detecting abnormal observations in the presence and absence of a training sample, respectively. It has been established that the use of ensembles of machine learning algorithms, as well as neural network algorithms, allows achieving high efficiency. anomaly detection. The results obtained are of practical importance for the development of systems for detecting and preventing cyberattacks in corporate networks, and also open up prospects for further improvement of network traffic protection technologies