Secure and Intelligent SDN Routing and Anomaly Detection Using XGBoost for Real-Time IoT Traffic Optimization

Abstract

The fast growth of the Internet of Things IoT technology generates extraordinary stress on network systems because it requires improved traffic performance and strengthened security measures. The central control capabilities of Software-Defined Networking (SDN) receive limited intelligence from conventional controllers when dealing with evolving network conditions and security threats. This study designs a protected and knowledgeable SDN routing framework for IoT traffic real-time optimization by integrating machine learning algorithms for path optimization and anomaly detection. The proposed method embeds two predictive models within the SDN controller: Light Gradient Boosting Machine (Light-GBM) for performance-aware routing optimization, and XG-Boost for real-time detection of malicious or anomalous flows. The system uses a hybrid decision-making pipeline for Quality of Service QoS measurement elements, such as latency, congestion level, and bandwidth utilization, together with security feedback like threat scores, blacklist status, and intrusion detection alerts. The system was tested by the research team using a simulated network infrastructure that emulates the common pattern of IoT traffic. The designed competitive metrics were latency, throughput, packet loss rate, accuracy of anomaly detection, false positive rate, and controller decision latency. Experimental findings suggest that the suggested SDN controller solution has a higher throughput and faster operations, with a difference to a baseline controller system of 42.7 and 49.7, respectively, and a 75.9% lower packet loss. The XG-Boost model gave an accurate detection rate of 99.8 percent with a false positive rate of 0.2 percent, with a controller decision time of 23.7 ms, which is compatible with real-time operations. Experimental evidence shows that the introduction of machine learning into SDN controller systems increases security operations in IoT and network routing implementation. This system demonstrates scalability together with modularity for direct implementation in operational SDN platforms, including POX and Ryu. The upcoming research will focus on implementing live network traffic while adapting models in real-time and deploying them in production IoT infrastructures.