AutoML-ID: automated machine learning model for intrusion detection using wireless sensor network.

Area of Science:

• Cybersecurity
• Machine Learning
• Wireless Sensor Networks

Background:

• Explainable machine learning models and synthetic data are increasingly popular for intrusion detection.
• Optimizing hyperparameters is crucial for model performance but often requires manual expertise.
• Existing hyperparameter tuning methods are problem-dependent and demand application-specific knowledge.

Purpose of the Study:

• To develop a cost-efficient machine learning model for fast intrusion detection and prevention in Wireless Sensor Networks (WSNs).
• To automate the selection of machine learning models and hyperparameter optimization using Automated Machine Learning (AutoML).
• To accurately predict the number of k-barriers for intrusion detection and prevention.

Main Methods:

• Utilized Automated Machine Learning (AutoML) with Bayesian optimization to select models and tune hyperparameters.
• Evaluated several explainable machine learning models including Gaussian process regression, support vector regression, and ensemble methods.
• Extracted synthetic predictors (area, sensing range, transmission range, number of sensors) via Monte Carlo simulation.
• Trained models on 80% of the data and tested on the remaining 20%.

Main Results:

• Gaussian process regression achieved exceptional performance, outperforming other models with R=1, RMSE=0.007, and bias=-0.006.
• AutoML demonstrated similar high performance when tested on a public intrusion detection dataset.
• The developed model accurately predicts the required number of k-barriers for intrusion detection.

Conclusions:

• Automated Machine Learning (AutoML) provides an effective solution for optimizing machine learning models in WSNs.
• Gaussian process regression is highly effective for intrusion detection and prevention tasks in WSNs.
• This research facilitates accurate prediction of k-barriers, enhancing intrusion detection capabilities.