Robust machine learning based Intrusion detection system using simple statistical techniques in feature selection

Area of Science:

• Cybersecurity
• Internet of Things (IoT)
• Industrial Control Systems

Background:

• Rapid expansion of IoT devices in Industry 4.0 presents significant security vulnerabilities.
• Resource-constrained IoT devices in challenging environments are susceptible to cyberattacks.
• Existing intrusion detection systems (IDS) face challenges in efficiency and effectiveness for IoT.

Purpose of the Study:

• To develop a lightweight intrusion detection system (IDS) for resource-limited IoT devices.
• To propose a novel feature selection algorithm to improve IDS performance and reduce computational overhead.
• To enhance the security of Industry 4.0 environments against cyber threats.

Main Methods:

• A unique feature selection algorithm utilizing basic statistical methods.
• Development of a lightweight intrusion detection system (IDS).
• Evaluation using IoTID20 and NSLKDD datasets with various classifiers.

Main Results:

• Reduced training time by 27-63% for multiple classifiers.
• Improved detection accuracy by selecting the most discriminative features.
• Achieved over 99.9% accuracy, precision, recall, and F1-Score on IoTID20 dataset.

Conclusions:

• The proposed lightweight IDS and feature selection effectively address security challenges in Industry 4.0 IoT.
• The methodology offers a significant improvement in performance and efficiency for IoT security.
• The system demonstrates robust and consistent performance across different datasets.