Purpose The emergence of Industry 4.0 has led to a substantial increase in the connectivity and integration of industrial technology and information systems. The impact of IoT and IIoT technologies is clearly demonstrated in the alteration of industrial processes and operations. A multitude of lightweight protocols have been devised to facilitate efficient communication in the context of the IoT. One of the protocols in this group is the MQTT protocol, also known as message queuing telemetry transfer. The goal of this research is to solve emerging threats in the IoT ecosystem and provide secure access to systems in the MQTT protocol through machine learning. Design This study adopts a multidimensional strategy, deploying several ML models in five scenarios, covering normal and attack instances, and fine-tuning them until they perform better. ML algorithms are developed and implemented by integrating them with the MQTT protocol for IoT communication. Methodology The methodology comprises researching and analyzing ML-driven methods (both classical and neural network models) using cross-validation performance indicators. Results Following the implementation of ten different models, the RNN, random forest, and decision tree models performed exceptionally well, with accuracy rates of 99.985%, 99.975%, and 99.96%, respectively. The files can be accessed at https://github.com/alisha2025/MQTTIntrusionDetection .