HDCBAN: Hybrid Neural Network for Network Intrusion Detection System

Intrusion Detection Systems (IDS) play a crucial role in safeguarding network security perimeters by monitoring network traffic and system behavior in real time. The advent of deep learning, particularly through the application of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, has significantly enhanced IDS's ability to detect network threats by leveraging autonomous learning and generalization capabilities. However, deep learning-based IDSs encounter challenges, including the necessity for extensive data labeling and the computational resources required, which pose obstacles to their widespread adoption and effectiveness. Nonetheless, deep learning-based IDS encounter challenges, including the requirement for extensive labeled datasets and significant computational resources. This paper proposes a novel hybrid IDS framework — HDCBAN that integrates deep convolutional neural networks, bidirectional LSTM networks, and the AlexNet model to efficiently predict and classify malicious network activities. The HDCBAN model employs deep CNNs to capture local features through convolution, bidirectional LSTMs to seize temporal features for enhanced performance and prediction capabilities, and AlexNet to augment classification efficacy through feature extraction. The effectiveness of this hybrid approach was assessed using real-world datasets, including the publicly available CSE-CIC-IDS 2017, CSE-CIC-IDS 2018, and NSL-KDD datasets, with preprocessing to optimize model performance. Experimental results, validated through 10-fold cross-validation, reveal that our model achieves a malicious attack detection rate and accuracy of up to 99.88% for CSE-CIC-IDS and 99.93% for NSL-KDD, thereby outperforming existing models in detection rate, accuracy, and reducing false positives.