A Hybrid Intrusion Detection Model Using LSTMAE and LightGBM for Robust Anomaly Detection in Network Systems

Traditional intrusion detection systems often face challenges such as low accuracy and high false positive rates, particularly when detecting ‎complex and evolving cyber threats. To address these limitations, a hybrid intrusion detection model is developed by integrating Long ‎Short-Term Memory Autoencoder (LSTMAE) with Light Gradient Boosting Machine (LightGBM). The LSTMAE component captures ‎temporal dependencies in network traffic, enabling effective feature extraction, while LightGBM performs efficient classification of the ex‎extracted features. The model is evaluated on benchmark datasets including NSL-KDD, UNSW-NB15, and CICIDS2017, following comprehensive preprocessing to ensure data consistency. Experimental results demonstrate that the hybrid model significantly outperforms ‎standalone classifiers and conventional methods, achieving improved detection accuracy and reduced false positive rates. These findings ‎highlight the model’s effectiveness in differentiating between normal and malicious traffic, offering a scalable and efficient solution for real-‎time intrusion detection, and laying the groundwork for future ensemble-based security frameworks‎.