Osprey optimization algorithm integrated with graph neural networks for intrusion detection in wireless sensor...

Area of Science:

• Cybersecurity
• Network Security
• Machine Learning

Background:

• Wireless Sensor Networks (WSNs) are increasingly vulnerable to sophisticated cyber-attacks.
• Conventional Intrusion Detection Systems (IDS) struggle with limited capabilities, high false alarm rates, and complex data processing.
• There is a critical need for advanced IDS to ensure WSN security.

Purpose of the Study:

• To propose a novel OOA-GNN model for enhancing intrusion detection in WSNs.
• To improve the accuracy and efficiency of attack detection while minimizing false alarms.
• To leverage deep learning on graphical structures for complex relationship identification in WSN data.

Main Methods:

• Developed a novel OOA-GNN framework integrating Osprey Optimization Algorithm (OOA) and Graph Neural Networks (GNN).
• Utilized a deep learning framework on graphical representations of WSN data to capture intricate network patterns.
• Applied Synthetic Minority Oversampling Technique (SMOTE) to address data imbalance in the Wireless Sensor Networks-Dataset (WSN-DS).
• Fine-tuned GNN hyperparameters using OOA to optimize detection performance.

Main Results:

• The OOA-GNN model achieved a high accuracy of 99.68% on the imbalanced WSN-DS dataset.
• OOA-GNN outperformed traditional classifiers including AdaBoost, GBM, XGBoost, KNN-AOA, and KNN-PSO.
• Demonstrated superior performance in terms of a low false positive rate and adaptability to network fluctuations compared to conventional methods.

Conclusions:

• The OOA-GNN model offers a significant advancement in WSN security through efficient and accurate intrusion detection.
• Integration of OOA's parameter-tuning with GNN's graph-based framework enhances real-time WSN operations.
• The proposed method effectively reduces false alarms and improves overall network reliability and attack detection precision.