A deep learning/machine learning approach for anomaly based network intrusion detection

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Network Security

Background:

• Cybersecurity threats are increasing in complexity and frequency.
• Advanced detection systems are needed for known and emerging attacks.
• Anomaly-based Network Intrusion Detection Systems (NIDS) are crucial for network defense.

Purpose of the Study:

• To develop a hybrid anomaly-based NIDS.
• To integrate multiple machine learning and deep learning algorithms.
• To improve the detection of diverse cybersecurity threats.

Main Methods:

• Utilized XGBoost, Random Forest, Graph Neural Networks (GNN), Long Short-Term Memory (LSTM), and Autoencoders.
• Trained on over 5.6 million network traffic records with extensive preprocessing.
• Employed Synthetic Minority Over-sampling Technique (SMOTE) and a weighted soft-voting ensemble strategy.

Main Results:

• Achieved near-perfect accuracy, precision, recall, and F1-scores on the primary dataset.
• Validated performance using 5-fold cross-validation.
• Demonstrated strong generalizability and robustness on an independent benchmark dataset.

Conclusions:

• The hybrid ensemble framework significantly enhances intrusion detection capabilities.
• The proposed NIDS is effective in complex and dynamic network environments.
• The system shows high potential for real-world cybersecurity applications.