Ensemble-Learning Framework for Intrusion Detection to Enhance Internet of Things' Devices Security

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• The Internet of Things (IoT) involves interconnected devices transmitting data, posing significant cybersecurity risks due to vulnerable network protocols.
• Cyber-attacks exploit these protocols, threatening the security and integrity of transmitted data.
• Existing Intrusion Detection Systems (IDS) require enhancement to effectively counter these evolving threats.

Purpose of the Study:

• To improve the detection efficiency of Intrusion Detection Systems (IDS) for Internet of Things (IoT) environments.
• To develop a robust method for binary classification of normal versus abnormal IoT network traffic.
• To evaluate the efficacy of ensemble machine learning classifiers in enhancing IDS performance.

Main Methods:

• Employed supervised machine learning algorithms including Random Forest, Decision Tree, Logistic Regression, and K-Nearest Neighbor.
• Utilized ensemble learning approaches (voting and stacking) to combine the strengths of individual classifiers.
• Trained and evaluated the proposed models on the TON-IoT network traffic dataset.

Main Results:

• Four individual supervised models (Random Forest, Decision Tree, Logistic Regression, K-Nearest Neighbor) demonstrated high accuracy.
• Ensemble classifiers (voting and stacking) achieved higher accuracy than individual models.
• The proposed framework achieved an overall accuracy rate of 0.9863 in improving IDS efficiency.

Conclusions:

• Ensemble learning strategies effectively enhance IDS reliability and reduce classification errors by leveraging diverse learning mechanisms.
• The developed framework significantly improves the efficiency and accuracy of Intrusion Detection Systems in IoT networks.
• The findings contribute to more secure data transmission within the Internet of Things.