Recent advances in pervasive technologies, such as wireless ad hoc networks and wearable sensor devices, allow the connection of everyday things to the Internet, commonly denoted as IoT (Internet of Things). IoT together with Big data technologies are seen as enablers to the development of intelligent and context-aware services and applications, e.g., healthcare, agriculture, and industrial systems. However, despite the efficiency of  lightweight communication protocols (e.g., MQTT), they are highly vulnerable to cyber-attacks, such as Distributed Denial of Service, Brute force, and SlowITe attacks. Recently, AI-driven solutions, particularly ML and DL, have shown significant potential in enhancing IoT security, especially for MQTT-based systems. However, datasets are required for evaluating those models in IoT-based attack detection. This work aims to provide such a dataset, named MQTTEEB-D, using a deployed IoT  Lab testbed,  with monitoring and processing capabilities. MQTTEEB-D is available both in its original form, allowing users to perform data preprocessing, and in a pre-processed format, for being used to further investigate the effectiveness of AI-driven predictive analytics for IoT security and attack prevention.