An Improved Network Intrusion Detection Method Based On CNN-LSTM-SA

• Main Author: Bian, Hui
• Format: Thesis
• Language: English; English
• Published: Universiti Malaysia Sarawak 2025
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://ir.unimas.my/id/eprint/47765/

As cyber threats grow in complexity and frequency, the importance of Network Intrusion Detection Systems in modern cybersecurity defense becomes increasingly critical. Traditional machine learning algorithms, such as Decision Trees, Naive Bayes, Random Forest, Random Trees, Multi-Layer Perceptron, and Support Vector Machines, have been extensively applied to address these threats. However, these algorithms often fall short in consistently detecting and classifying network intrusions, particularly when distinctions between classes are subtle or when facing evolving attack patterns. To overcome these limitations, this research proposes the CNN-LSTM-SA method, an enhanced deep learning approach that integrates Convolutional Neural Networks, Long Short-Term Memory networks, and Self-Attention mechanisms. This integration optimizes feature extraction by capturing both spatial and temporal relationships, enhancing the detection of complex network behaviors. Using the NSL-KDD dataset for evaluation, the proposed method demonstrates superior performance compared to conventional algorithms and related deep learning techniques, achieving higher precision, recall, F1 scores and overall accuracy in both binary and multi-class classification tasks. Furthermore, comparative analysis with existing methods in the field highlights the CNN-LSTM-SA model's ability to address scalability, adaptability and classification challenges. By leveraging deep learning techniques, the CNN-LSTM-SA method demonstrates notable potential in enhancing cybersecurity measures against evolving threats.