Sistem Deteksi Intrusi Menggunakan Teknik Undersampling Enn dan machine learning Random forest untuk Mengatasi Ketidak seimbangan

Data imbalance in intrusion detection systems can cause the model to overlook attack types with limited data samples. This study develops an intrusion detection system based on the Random Forest algorithm with a data balancing approach using Edited Nearest Neighbours (ENN) on the CICIDS2017 dataset. The research stages include preprocessing, normalization, feature selection, and the application of balancing methods. Initial results indicate that the use of ENN alone does not fully balance the class distribution. Therefore, it is recommended to implement a hybrid SMOTE-ENN method, which has proven more effective in improving class distribution and enhancing the detection of minority attacks. Model evaluation shows a significant performance improvement, with accuracy reaching 98.66% and precision, recall, and F1-score values exceeding 98% across various data-splitting scenarios.