Network-based intrusion detection using deep learning technique.

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Rapid network traffic growth and sophisticated cyber threats necessitate advanced intrusion detection systems.
• Traditional Network-based Intrusion Detection Systems (NIDS) struggle with novel attack patterns due to outdated data and conventional machine learning models.
• Existing NIDS face limitations in detecting evolving threats, highlighting the need for more robust and adaptive solutions.

Purpose of the Study:

• To introduce a novel deep learning approach for enhanced network intrusion detection.
• To address the limitations of traditional NIDS by integrating Sequential Deep Neural Networks (DNN) with Rectified Linear Unit (ReLU) activation and Extra Tree Classifier for feature selection.
• To improve the accuracy, interpretability, and computational efficiency of NIDS for real-time threat detection.

Main Methods:

• Developed a hybrid model combining Sequential Deep Neural Networks (DNN) with Rectified Linear Unit (ReLU) activation.
• Employed the Extra Tree Classifier for optimizing feature selection, reducing the feature space from 43 to 8 highly relevant features.
• Trained and validated the model using the comprehensive UNSW-NB15 dataset, simulating realistic network traffic and attack vectors.

Main Results:

• The proposed Sequential DNN model achieved high performance metrics: 97.93% accuracy, 97% Precision, 97% Recall, and 97% F1-score for binary classification (normal vs. attack).
• Feature optimization via Extra Tree Classifier significantly reduced computational load and improved inference speed without compromising accuracy.
• Experimental validation, including ROC curves and Confusion Matrices, demonstrated superior performance compared to existing studies.

Conclusions:

• The integration of ReLU-based DNN with optimized feature selection offers a powerful and interpretable solution for network intrusion detection.
• The proposed model effectively overcomes challenges like vanishing gradients and overfitting, crucial for reliable NIDS deployment.
• This advanced NIDS presents a promising solution for enhancing cybersecurity in critical infrastructure sectors like finance, healthcare, and government.