A Genetic-Based Extreme Gradient Boosting Model for Detecting Intrusions in Wireless Sensor Networks

Area of Science:

• Computer Science
• Network Security
• Machine Learning

Background:

• Wireless Sensor Networks (WSNs) are vulnerable to diverse and unpredictable cyber threats.
• Existing machine learning intrusion detection systems struggle with imbalanced network traffic data.
• Improved accuracy and efficiency are crucial for robust WSN security.

Purpose of the Study:

• To propose a novel intrusion detection model, GXGBoost, for WSNs.
• To address the challenge of imbalanced data in WSN network traffic.
• To enhance the detection of minority attack classes.

Main Methods:

• Developed the GXGBoost model, integrating a genetic algorithm with an extreme gradient boosting (XGBoost) classifier.
• Utilized the wireless sensor network-detection system (WSN-DS) dataset for experiments.
• Employed holdout and 10-fold cross-validation techniques to evaluate model performance.

Main Results:

• The GXGBoost model demonstrated superior performance compared to state-of-the-art methods.
• Achieved high detection rates: 98.2% for flooding, 92.9% for scheduling, 98.9% for grayhole, and 99.5% for blackhole attacks.
• Attained an excellent 99.9% detection rate for normal traffic.

Conclusions:

• The proposed GXGBoost model effectively detects intrusions in WSNs, particularly in imbalanced datasets.
• The integration of genetic algorithms and XGBoost offers a powerful solution for WSN security.
• GXGBoost significantly advances the state-of-the-art in intrusion detection for wireless sensor networks.