IoT setups face unique problems like limited computing power, different protocols, and varied devices, making old detection methods less effective. This study presents a new machine learning way to spot DDOS attacks in networks. Our mixed model joins anomaly detection and supervised learning to boost accuracy and speed. This approach uses an anomaly detection system to find odd network traffic patterns and a supervised learning method to sort them as safe or dangerous. We test the proposed system using big datasets covering many DDOS attack types and RT_IOT22, which looks at IoT-specific network and attack methods. Our tests show that our method improves IoT defence against DDOS attacks. The results suggest that the mixed approach fits well in the real world. We are offering a trusted and scalable way to enhance network safety.