Studi Optimalisasi Deteksi Intrusi Jaringan NIDS Menggunakan XGBoost pada Dataset Netflow V2

Aritonang, Mhd Adi Setiawan
Marshall Al Karim, Muhammad
Roland, Roland
Irwan Gultom, Jefri
Enrico Sitompul, muel
Hendri , Hendri

Jurnal Pustaka AI : Pusat Akses Kajian Teknologi Artificial Intelligence
Vol 6 No 1 (2026): Pustaka AI (Pusat Akses Kajian Teknologi Artificial Intelligence)

This research is motivated by the increasing complexity of cyber attacks on modern networks, necessitating the need for an adaptive and accurate network intrusion detection system (NIDS) through the use of machine learning algorithms, specifically XGBoost. This research uses the NF-UQ-NIDS-v2 dataset with structured pre-processing stages, stratified data partitioning, and the development of an XGBoost-based multi-class classification model with optimized hyperparameter configurations. The test results show that the XGBoost model achieves an overall accuracy of 98.84% with excellent performance in the majority class, but still experiences a decrease in performance in the minority class due to data imbalance. The XGBoost-based NIDS model is proven to be effective and stable in detecting large-scale network attacks, although further strategies are needed to improve detection capabilities for rare types of attacks..