Analysis of Machine Learning Methods for Detecting Computer Attacks

In the context of constantly evolving threats in the field of cybersecurity, the detection of new attack classes is a critically important task. This study examines machine learning methods, including neural networks, as classifiers for detecting computer attacks. In order to analyze the effectiveness of attack detection, experiments were conducted on modern datasets. The dataset includes classes of computer attacks, as well as architectures of classifiers such as: multilayer perceptron, deep belief network, sparse and deep autoencoder with catboost, recurrent neural networks. The conducted experiments demonstrate the effectiveness of machine learning methods and their ability to detect new attack classes.