Improving Intrusion Detection System by Estimating Parameters of Random Forest in Boruta

To overcome the security problem of computer networks, the Intrusion Detection System (IDS) is developed. It is intended to identify an attack. Various types of IDS are built according to the environment: signature-based and anomaly-based. This second type of IDS can identify attacks that have not been known. In this case, machine learning is a possible method to develop an IDS model, which comprises many processes, including feature selection. The Boruta Algorithm is a feature selection method that is good enough to apply to machine learning. However, in its application on the NSL-KDD dataset, this algorithm has an infinite loop problem. This paper presents the analysis and estimation of random forest parameters, precisely the depth and number of trees; additionally, the use of entropy and Gini index as z-score in the Boruta Algorithm is considered. The experimental result shows that the proposed method can prevent the infinite loop, which indirectly improves the performance of the existing algorithm.