Cyberattacks Classification by Tuning Deep Hyperparameters Using Bayesian Optimization

With the rapid development of the Internet, cyber-attacks have become more frequent, making network intrusion detection systems (NIDSs) essential for maintaining system security. Recently, there has been a significant focus on artificial intelligence (AI)-based anomaly detection systems within NIDSs, leading to the proposal of various models aimed at enhancing their performance. However, these AI models often face challenges due to data imbalance, where insufficient learning of malicious behaviors results in inaccurate threat detection. This paper presents a deep learning (DL) approach to NIDS that employs the synthetic minority over-sampling technique (SMOTE) to address the data imbalance issue, which achieves exceptional performance in both binary and multi-class classification tasks. Our proposed system leverages a residual connection convolutional neural network (CNN). To validate its effectiveness, we conducted experiments using the NSL-KDD and CIC-IDS2017 datasets. The results demonstrate that our model achieves a multi-class classification accuracy of up to 91.40% on the NSL-KDD dataset and 99.53% on the CIC-IDS2017 dataset. For binary classification, the accuracy reaches up to 94.15% on the NSL-KDD dataset and 99.61% on the CIC-IDS2017 dataset.