Network Attack Detection Using Machine Learning Methods

Abstract

This paper presents the result of the study of network intrusion detection using machine learning algorithms. The creation and training of such algorithms is seriously limited by the small number of actual datasets available for public access. The CSE-CIC-IDS2018 data set, used in research, includes 7 subsets of different attack scenarios. Each subset is labeled using a few subtypes of a given attack or normal behavior. That is why the problem of network attack detection has been considered a multiclassification problem. Some of the most popular classifiers will be tested on the chosen data set. Classification algorithms are developed using a standard Python programming environment and the specialized machine learning library Scikit-learn. In the paper, a comparative analysis of the results was performed based on the the application of Random Forest, XGBoost, LR, and MLP classifiers.