Shallow Learning Techniques for Early Detection and Classification of Cyberattacks over MQTT IoT Networks

Area of Science:

• Cybersecurity
• Network Security
• Internet of Things (IoT)

Background:

• Global connectivity via the Internet of Things (IoT) increases system vulnerabilities.
• Cyberattackers exploit IoT device limitations and wireless vulnerabilities for network compromise.
• Existing security measures struggle with resource-constrained IoT environments.

Purpose of the Study:

• To develop a shallow learning multiclassifier for detecting and classifying cyberattacks on IoT networks.
• To specifically address Denial-of-Service (DoS) and Intrusion attacks in MQTT-based IoT systems.
• To enable efficient, local network monitoring on resource-constrained IoT devices.

Main Methods:

• A shallow learning multiclassifier approach was employed.
• Inter-device communication data from MQTT networks was utilized for attack detection.
• The system was designed for implementation on resource-constrained IoT devices.

Main Results:

• The system achieved over 99% accuracy in detecting cyberattacks.
• An F1-score greater than 80% was obtained for Intrusion attack detection.
• Validated on a real-world IoT dataset using MQTT communication.

Conclusions:

• Shallow learning multiclassifiers are effective for IoT cybersecurity.
• The proposed system enhances security and incident response capabilities in IoT networks.
• Local network monitoring on resource-constrained devices is feasible and beneficial.