Fault Detection in Wireless Sensor Networks through the Random Forest Classifier

Area of Science:

• Computer Science
• Electrical Engineering
• Network Security

Background:

• Wireless Sensor Networks (WSNs) face significant fault vulnerabilities due to unpredictable environments, leading to hardware, software, and communication failures.
• Limited sensor resources and diverse deployment scenarios complicate fault detection in WSNs, necessitating robust methodologies.
• Sensor-level fault classification is crucial for maintaining the integrity and performance of WSNs.

Purpose of the Study:

• To conduct a comparative analysis of various machine learning classifiers for detecting sensor-level faults in Wireless Sensor Networks.
• To evaluate the effectiveness of Support Vector Machine (SVM), Convolutional Neural Network (CNN), Stochastic Gradient Descent (SGD), Multilayer Perceptron (MLP), Random Forest (RF), and Probabilistic Neural Network (PNN) for fault classification.
• To identify the most accurate and reliable algorithm for fault detection in WSNs.

Main Methods:

• Utilized six distinct machine learning classifiers: SVM, CNN, SGD, MLP, RF, and PNN.
• Focused on classifying sensor-level faults including gain, offset, spike, data loss, out of bounds, and stuck-at faults.
• Induced spike and data loss faults into real-world datasets for empirical evaluation.

Main Results:

• Compared classifier performance using metrics such as Detection Accuracy (DA), True Positive Rate (TPR), Matthews Correlation Coefficients (MCC), and F1-score.
• Simulations indicated that the Random Forest (RF) algorithm achieved a superior fault detection rate compared to other evaluated classifiers.
• RF demonstrated higher accuracy in identifying induced faults like spike and data loss.

Conclusions:

• The Random Forest (RF) algorithm is highly effective for sensor-level fault detection in Wireless Sensor Networks.
• Comparative analysis highlights RF's advantage over other tested machine learning models in terms of fault detection accuracy.
• The findings suggest RF as a promising solution for enhancing the reliability and robustness of WSNs against common sensor failures.