This work addresses an approach for generating a dataset collection based on an executive review of attacks on application layer protocols of the IoT environments, such as CoAP and MQTT. In the smart cities, the IoT devices have a significant role and their security becomes very important. Due to the fact that special characteristics of IoT devices in terms of processing power, energy savings and cost, makes them a more heterogeneous method of communication than conventional networks. The use of these protocols raises new challenges in the cybersecurity. In order to address security without increasing the complexity of IoT systems, the network traffic must be analyzed in deep with machine learning techniques. Therefore, network traffic datasets are of high importance to detect threats through a dynamic intrusion detection system based on intelligent models. So, the system implemented and shows in this paper is a viable way to generate dataset where attacks described are included.