The rapid expansion of Internet of Things (IoT) devices in smart home ecosystems has significantly elevated concerns over network security, particularly in relation to Distributed Denial-of-Service (DDoS) attacks. These threats are intensified by the diversity in device capabilities and the limited computational resources typical of household systems. Existing security infrastructures, which often depend on uniform traffic analysis and centralized cloud-based mitigation strategies, fall short in addressing the unique behavioral patterns and vulnerabilities of heterogeneous IoT environments. In response to these challenges, this study introduces SDN-OvR, a novel framework that integrates Software-Defined Networking (SDN) with One-vs-Rest (OvR) machine learning classification. Through SDN's centralized and programmable control capabilities, the proposed approach dynamically identifies and profiles individual IoT devices, such as surveillance cameras and environmental sensors, enabling tailored anomaly detection. Device-specific Support Vector Machine (SVM) models are trained to accurately distinguish between benign and malicious traffic, achieving a classification accuracy of 98.7% while simultaneously lowering false positives by 32% relative to traditional models. The SDN-OvR framework further incorporates a real-time mitigation engine, which leverages OpenFlow protocols to enforce security policies with an average response latency of 13.2 milliseconds—delivering threefold performance gains over conventional platforms like Cisco Stealthwatch. Validation of the system was carried out using both the CICDDoS2019 dataset and a purpose-built smart home testbed comprising over 50 devices. Experimental results confirmed its scalability to networks exceeding 1,000 nodes, maintaining processing overhead below 10% CPU utilization. Noteworthy contributions of this work include the design of a novel feature engineering pipeline tailored to extract 12 IoT-specific traffic features, an open-source release incorporating the newly developed IoT-DDoS-2023 dataset, and comprehensive quality-of-service (QoS) evaluation under varying threat conditions. By aligning intelligent traffic management with adaptive defense strategies, the SDN-OvR framework presents a viable, deployable solution for enhancing DDoS resilience in residential and small-scale enterprise IoT environments.

Keywords: Software-Defined Networking (SDN), IoT Security, DDoS Mitigation, One-vs-Rest Classification, Machine Learning, Smart Home