Comparison of Random Forest, K-Nearest Neighbor, and Support Vector Machine Classifiers for Intrusion Detection System

Earlier classification investigations identified Random Forest (RF), k-Nearest Neighbor (kNN), and Support Vector Machine (SVM) as leading nonparametric classifiers capable of achieving high accuracies. Nonetheless, there has been limited research comparing the performances of these classifiers across various standard intrusion detection datasets. In this research, the performances of the RF, KNN, and SVM classifiers are examined and compared when they are used for intrusion detection on various standard intrusion detection datasets. For each of the algorithms, parameter tunings produced various accuracies on each of the datasets. The parameter with the highest accuracy for each algorithm is compared with its counterpart in other algorithms on the same dataset. The experimental results show that the KNN Intrusion Detection System outperformed other approaches on all the datasets, with accuracies of 0.999928556, 0.996921593 and 0.971580496 on NSL-KDD, CICIDS2017 and CICIDS2018 datasets, respectively. Consequently, the obtained results of KNN present better performances in terms of precision, recall and f1-score on the various network traffic classes when compared to the other algorithms. As shown in this research, the high performances of these machine learning algorithms show that they can be deployed in the field for intrusion detection.