The integration of Software-Defined Networking (SDN) with the Internet of Things (IoT) provides significant advantages such as centralized control, enhanced scalability, and more efficient network management. However, this convergence introduces substantial security challenges due to the expanded attack surface created by the massive and heterogeneous data generated by IoT devices. Traditional Intrusion Detection Systems (IDS) often struggle to effectively analyze and manage such dynamic traffic, leading to reduced detection accuracy and increased vulnerability to emerging threats. To address these challenges, this paper proposes a machine learning–based ensemble IDS specifically designed for SDN-enabled IoT environments. The proposed system , implemented in Python 3, employs a hard voting strategy that combines Random Forest, Decision Tree, and K-Nearest Neighbors (KNN), with feature selection applied individually to each classifier to improve performance before ensembling. Experimental evaluation conducted using the NSL-KDD dataset demonstrates that the proposed approach achieves an accuracy of 99.67%, precision of 99.78%, recall of 99.53%, and F1-score of 99.65%. Additional validation on the UNSW-NB15 dataset further confirmed the robustness and generalizability of the method. These results indicate the system’s high effectiveness in detecting diverse attack patterns with minimal false alarms, making it suitable for real-world SDN-IoT environments.