Network Intrusion Detection System Based on Ensemble Learning Using CIC-IDS2017

Abstract

Increasingly complex cyber threats demand accurate and efficient intrusion detection. This study develops an ensemble of Random Forest, XGBoost, and LightGBM to detect attacks on CIC-IDS2017. SMOTE addresses class imbalance. The ensemble achieves a 98.1% F1-score with average inference of 12 ms per flow, making it suitable for enterprise network environments.