Attack-Aware IoT Network Traffic Routing Leveraging Ensemble Learning.

Area of Science:

• Cybersecurity
• Machine Learning
• Internet of Things (IoT)

Background:

• Network Intrusion Detection Systems (NIDSs) are crucial for defending against cyberattacks.
• Anomaly-based NIDSs build behavioral profiles for detecting threats in IoT networks.

Purpose of the Study:

• To design, implement, and evaluate machine-learning-based NIDS for IoT environments.
• To compare the performance of ensemble, neural network, and kernel methods for NIDS.

Main Methods:

• Evaluated six supervised learning algorithms across three classes: ensemble, neural network, and kernel methods.
• Utilized distilled-Kitsune-2018 and NSL-KDD datasets containing real-world IoT network traffic with attacks.
• Applied standard machine learning metrics to assess identification accuracy, error rates, and inference speed.

Main Results:

• Ensemble methods demonstrated higher accuracy and lower error rates compared to neural network and kernel methods.
• Neural network methods achieved the highest inference speed, suitable for high-bandwidth networks.
• Achieved performance improvements of 1-20% over existing state-of-the-art NIDS solutions.

Conclusions:

• Machine learning, particularly ensemble methods, significantly improves NIDS performance in IoT networks.
• The choice between ensemble and neural network methods depends on whether accuracy or speed is prioritized.
• The developed NIDS models offer a substantial advancement in detecting network intrusions in IoT environments.