AI/IoT combination have adopted in various corporate fields as the trigger of the 4th revolution over the next several years, followed by the ongoing advancement of both IoT (Internet of Things) and AI (Artificial Intelligent) technology. Machine Learning (ML), software development. IoT may benefit greatly from the combination of AI and Big Data analytics. In cybersecurity, the confidentiality, availability and integrity are widely recognized principles. Various attacks from different external or internal sources are revealed in an IoT network. An Intrusion Detection System (IDS) is needed to deal with attacks through IoT networks. An IDS is a device, physical device or software application, which detects activity of any malicious on a network and triggers an alert so that IDSs act as watchful eyes. Machine learning, a component of AI, is a modern technology that is used to improve the quality and IDSs accuracy. This paper’s aim is to build an improved IDS using the proposed ensemble machine learning (CNN-DTXG) and examined it against the existing machine learning (DT, xgboost, LR, kNN, RF, NB, CATBOOST) in order to improve IoT networks against cyber attacks for two datasets (CIC_IDS2023 and IoTID20) in terms of Accuracy, Precision, FI-Score, Recall and ROC-AUC (Region of Convergence—Area Under the Curve). In this paper, CIC-IoT2023 dataset has been combined with Ransomware dataset, representing one of the critical IoT cyber attacks so that the expected results showed that the proposed ensemble model had improved accuracy and reliability with reached to 95.034% and 99.630% for multiple and binary classification for IoTID2023 and the accuracy and reliability reached to 99.99% for IoTID20 for multiple and binary classification indicating better improvement for the performance of the system. The results also showed that the area under the curve AUC reached 0.95025 for CIC-IoT2023 and 0.99994 for IoTID20 resulting in better performance as AUC reached to 1.0.