The explosive growth of the domain of the Internet of things (IoT) network devices has resulted in unparalleled ease of productivity, convenience, and automation, with Message Queuing Telemetry Transport (MQTT) protocol being widely recognized as an essential communication standard in IoT environments. MQTT enables fast and lightweight communication between IoT devices to facilitate data exchange, but this flexibility also exposes MQTT to significant security vulnerabilities and challenges that demand highly robust security. This paper aims to enhance the detection efficiency of an MQTT traffic intrusion detection system (IDS). Our proposed approach includes the development of a binary balanced MQTT dataset with an effective feature engineering and machine learning framework to enhance the security of MQTT traffic. Our feature selection analysis and comparison demonstrates that selecting a 10-feature model provides the highest effectiveness, as it shows significant advantages in terms of constant accuracy and superior training and testing times across all models. The results of this study show that the framework has the capability to enhance the efficiency of an IDS for MQTT traffic, with more than 96% accuracy, precision, recall, F1-score, and ROC, and it outperformed the most recent study that used the same dataset.
Keywords: Internet of things; Message Queuing Telemetry Transport; feature selection; machine learning; security.