The widespread adoption of IoT devices has revolutionized multiple sectors, including healthcare, military, agriculture, and smart cities. This surge in IoT-generated data raises significant security concerns, necessitating efficient strategies for large-scale data analysis to safeguard IoT devices. Existing research has explored the fusion of Software-Defined Networking (SDN) and machine learning (ML), particularly flow-based monitoring, for intrusion detection. However, as IoT data volumes grow, challenges such as scalability, adaptability to new attack vectors, and resource-intensive monitoring persist. Our solution combines SD-IoT and ML to enhance IoT network security. By isolating virtual networks based on device characteristics, we improve intrusion detection efficiency and facilitate research on emerging threats. We present a real-world implementation, demonstrating a scalable and robust ML-based security for SD-IoT system.