Enhancing Network Intrusion Detection: A Comparative Analysis of Machine Learning Models on Complex Network Traffic Data
Jie Hu, Xiaohan Wang, Yuan Liu · 2024
The exponential growth of IoT and cloud technologies presents new challenges for network security, requiring more robust intrusion detection systems (IDS). Current systems struggle with scalability and accuracy, underscoring the need for advanced solutions. This study evaluates the performance of machine learning models—LSTM, Random Forest, Isolation Forest, GBM, and XGBoost—on a network intrusion detection task using a Kaggle dataset. GBM and XGBoost outperformed other models, demonstrating superior accuracy, recall, and F1 scores, particularly in detecting minority class instances. Random Forest also performed well but was outpaced by the precision and robustness of GBM and XGBoost. The results emphasize the need for advanced data balancing and ensemble techniques to enhance detection accuracy, offering a pathway to more scalable and resilient IDS.