Method for Selecting Features to Analyze The Performance of Intrusion Detection Systems

Intrusion prevention systems (IPS) and intrusion detection systems (IDS) in place for computer networks is essential for a company to prosper. These components play a vital role in guaranteeing the company's success. In order to guarantee that computer networks within businesses are safe, dependable, and accessible, several strategies have been used in the development and implementation of IDSs and IPSs over the years. Constructed using machine learning techniques are IDSs that deserve special attention. IDSs that use machine learning techniques are precise and successful at identifying network assaults. However, for high dimensional data spaces, these systems' performance degrades. As a result, it is essential to use a suitable feature extraction technique it is feasible to get remove attributes that do not have a significant impact on the classification process. Furthermore, when trained on highly imbalanced datasets, numerous ML-based intrusion detection systems experience an increase in false positive rates and a decline in detection accuracy. In order to implement the following machine learning techniques using the reduced feature space, will also be applied using the XGBoost algorithm. The techniques involved in this analysis include Support Vector Machine (SVM), kNearest-Neighbor (kNN), Logistic Regression (LR), Artificial Neural Network (ANN), and Decision Tree (DT). The study considered both binary and multi-class classification scenarios. The results showed that the use of methods such as DT together with the XGBoost-based feature selection method can improve the test accuracy of the binary classification system.