Machine Learning based Intrusion Detection System for IoT Applications using Explainable AI

This research focuses on studying the classification performance of a Machine Learning-based Intrusion Detection System (IDS) using the UNSW-NB15 dataset. The effectiveness of three classifiers - Decision Tree, Multilayer Perceptron (MLP), and XGBoost - was analyzed to determine their accuracy in identifying attacks and normal network traffic. The experimental results revealed that Decision Tree achieved an accuracy of 96.5%, MLP achieved an accuracy of 89.83%, and XGBoost achieved an accuracy of 89.9%. Additionally, the Explanability of the machine learning models was analyzed, highlighting the differences in interpretability among the classifiers. It was observed that Decision Tree provided better Explanability, but lower accuracy compared to MLP and XGBoost. Overall, this research contributes to our comprehension of the performance and Explanability of three different machine learning classifiers for intrusion detection. The findings can provide valuable insights for choosing suitable classifiers that align with the specific priorities and requirements of the IDS system.