INTERPRETABLE PREDICTIVE MODEL OF NETWORK INTRUSION USING SEVERAL MACHINE LEARNING ALGORITHMS

Abstract

Network intrusion is any unauthorized activity on a computer network. Attacks on the network computer system can be devastating and affect networks and company establishments. Therefore, it is necessary to curb these attacks. Network Intrusion Detection System (NIDS) contributes to recognizing the attacks or intrusions. This paper explains the factors that influence network attacks. Some machine learning methods are used such as are logistic regression, random forest XGBoost, and CatBoost. The best model is chosen from these models based on its accuracy level. Classification modelling is divided into two types, namely using a dummy and not using dummy variables. The best method for predicting network intrusion is a random forest with a dummy variable that has an Area Under Curve (AUC) value of 92.31% and an accuracy of 90.38%.