AIoT (Artificial Intelligence of Things) is a modern word that has emerged as a significant issue, integrating two significant acronyms: AI (Artificial Intelligence) and IoT (Internet of Things). A critical security requirement had been the necessity for a more robust cybersecurity framework to oversee and mitigate risks associated with data. In order to maintain a robust cybersecurity defense and mitigate the risks posed by cyberattacks on their infrastructure, several corporations have integrated Artificial Intelligence (AI) into their threat intelligence systems. The deployment of several AI techniques has been crucial and increasingly utilized to improve the performance of Intrusion Detection Systems (IDS). IDS monitors a network, a host for security violations and alert the administrator upon detection. Updating the network with a sample of the attack type is crucial for its detection capabilities. This paper aimed to built an IDS utilizing a proposed hybrid model and examined it against the traditional deep learning models like (DNN, LSTM, GRU, DAENN and RNN). This study is investigated for three datasets (IoTID20, CICDDOS_2019 and CICDDOS_2017) which contains different types of IoT attacks. The results outperformed the previous models by achieving accuracy and reliability reached to 99.91%, 99.35% and 96.90% for IoTID20, CICDDOS_2019 and CICDDOS_2017 respectively.