A hybrid intrusion detection system for an imbalanced dataset using deep learning

Network intrusion detection is crucial for detecting devastating internet attacks. The expanding amount of data from many sources is a worry in the absence of an intrusion detection system. Several studies have been carried out in order to detect malicious attack types. Because the intrusion detection system must cope with imbalanced data in the real world, a model’s overall accuracy is insufficient. As a result, minor class event identification with higher accuracy is crucial in addition to overall accuracy. Deep learning surpasses traditional machine learning techniques in the majority of research. In this study, we described a hybrid deep learning model that integrates convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM). CNN-BiLSTM is the name of the model, which includes spatial and temporal data learning. We employed data level modification approaches such as the Synthetic Minority Over-sampling Technique (SMOTE) and Adaptive Synthetic (ADASYN) to address the data imbalance. To test our proposed model, we used binary and multi-class datasets. For binary datasets, we used the Imblearn Python module, while for multi-class classification, we used the NSL-KDD dataset. Not only does our proposed data level modification during preprocessing improve overall model accuracy, but it also improves minor class accuracy.