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

Abstract

In this paper, we introduce the framework and its experimental design, which was used to elaborate the MQTTEEB-D dataset and to execute real-time MQTT-based attacks while collecting traffic data. The MQTTEEB-D dataset is a practical real-world data set for intrusion detection improvement in MQTT-based IoT networks. Unlike already existing datasets, which are constructed using simulated network traffic, MQTTEEB-D is obtained from a real-time IoT testbed, named MQTTEEB. Various cyberattacks, including Denial of Service (DoS), Slow DoS against Internet of Things Environments (SlowITe), Malformed Data Injection, Brute Force, and MQTT publish flooding, were carried out in real-time while allowing close monitoring of network traffic anomalies. Data was gathered using PyShark and organized into multiple CSV files. To ensure high data quality, we performed pre-processing steps, such as outlier removal, normalization, standardization, and class balance. Several processed forms of data (e.g., raw, cleaned, normalized, standardized), along with detailed metadata are also provided for being used, for instance, to develop and validate AI-driven intrusion detection models.