MQTTEEB-D: A Real-World IoT Cybersecurity Dataset for AI-Powered Threat Detection in MQTT Networks

Area of Science:

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

Background:

• Existing IoT datasets often rely on simulated traffic, limiting their real-world applicability.
• MQTT is a prevalent messaging protocol in IoT, making its security a critical concern.
• Effective intrusion detection systems (IDS) require realistic, diverse datasets for accurate model training and validation.

Purpose of the Study:

• To introduce the MQTTEEB-D dataset, a novel, real-world dataset for MQTT-based IoT network security research.
• To provide a comprehensive collection of real-time MQTT-based cyberattack traffic data.
• To facilitate the development and validation of advanced AI-driven intrusion detection models for IoT environments.

Main Methods:

• Development of the MQTTEEB framework and experimental design for data collection.
• Execution of various real-time cyberattacks (e.g., DoS, SlowITe, Brute Force, MQTT publish flooding) on an IoT testbed.
• Data acquisition using PyShark, followed by rigorous pre-processing including outlier removal, normalization, standardization, and class balancing.

Main Results:

• The MQTTEEB-D dataset was successfully generated from a real-time IoT testbed (MQTTEEB).
• The dataset captures diverse network traffic anomalies resulting from multiple, realistic cyberattacks.
• Multiple processed versions of the dataset, including raw, cleaned, normalized, and standardized data, are available.

Conclusions:

• The MQTTEEB-D dataset offers a valuable, practical resource for advancing intrusion detection in MQTT-based IoT networks.
• The dataset's real-world nature and comprehensive attack scenarios enable more robust AI model development.
• This work contributes to enhancing the security posture of interconnected IoT systems against sophisticated cyber threats.