摘要
The fast development of Internet technologies ignited the growthof techniques for information security that protect data, networks, systems,and applications from various threats. There are many types of threats. Thededicated denial of service attack (DDoS) is one of the most serious andwidespread attacks on Internet resources. This attack is intended to paralyzethe victim’s system and cause the service to fail. This work is devoted tothe classification of DDoS attacks in the special network environment calledSoftware-Defined Networking (SDN) using machine learning algorithms. Theanalyzed dataset included instances of two classes: benign and malicious.As the dataset contained twenty-two features, the feature selection techniques were required for dimensionality reduction. In these experiments, theInformation gain, the Chi-square, and the F-test were applied to decreasethe number of features to ten. The classes were also not completely balanced, so undersampling, oversampling, and synthetic minority oversampling(SMOTE) techniques were used to balance classes equally. The previousresearch works observed the classification of DDoS attacks applying variousfeature selection techniques and one or more machine learning algorithms.Still, they did not pay much attention to classifying the combinations offeature selection and balancing methods with different machine learningalgorithms. This work is devoted to the classification of datasets with eightmachine learning algorithms: naïve Bayes, logistic regression, support vectormachine, k-nearest neighbors, decision tree, random forest, XGBoost, andCatBoost. In the experimental results, the Information gain and F-test featureselection methods achieved better performance with all eight ML algorithmsthan with the Chi-square technique. Furthermore, the accuracy values of theoversampled and SMOTE datasets were higher than that of the undersampledand imbalanced datasets. Among machine learning algorithms, the accuracyof support vector machine, logistic regression, and naïve Bayes fluctuatesbetween 0.59 and 0.75, while decision tree, random forest, XGBoost, and CatBoost allowed achieving values around 0.99 and 1.00 with all featureselection and class balancing techniques among all the algorithms.
