Ensemble Learning and Hybrid Deep Learning Based Intrusion Detection with XAI Insights

Network security has emerged as a critical concern in the context of the increased connectivity of devices and the significance of safeguarding network data. To address this issue, the development of a robust and reliable system is imperative. This study implements a network intrusion detection system on the NSL-KDD benchmark dataset to monitor and identify malicious activities within the network. Six supervised machine learning algorithms, including ensemble learning techniques, are utilized for multiclass classification. Additionally, a combination of neural network models employing convolutional neural network, long short-term memory (LSTM) and their combined architectures are applied to the employed open-source dataset. Extensive preprocessing and feature extraction techniques are employed to extract salient features, enabling the models to achieve optimal training outcomes. These approaches facilitate the detection and classification of network attacks. Among the various classification models evaluated, the XGBoost classifier demonstrates the highest accuracy of 98.63%, closely followed by the Bi-LSTM model, achieving an accuracy of 97.83%. An explainable AI approach utilizing the LIME framework is employed to gain insights into the model’s decision-making process.