Enhancing IOT-SDN Integration with Deep Learning for Network Attack Mitigation Using Residual Yolov7 Approach

Abstract

The Internet of Things (IoT) and Software-Defined Networking (SDN) are changing current networking with better flexibility and management. However, also create new security problems, particularly because hackers targeting SDN infrastructure often utilise IoT devices as entry points. IoT-SDN configurations are exposed to a wide range of security threats because of their complexity and increasing interconnectedness. We propose a new approach to enhance network security in IoT-SDN ecosystems by means of deep learning, more precisely the Residual YOLOv7 framework. Because residual learning records and adjusts temporal data, YOLOv7 can precisely identify anomalies and attacks. By use of real-time analytics, this method continuously monitors network traffic, detects anomalous activity, and responds quickly to any threats. Strong network defence and efficient resource allocation are facilitated by the precise attack identification made feasible by the inclusion of Residual YOLOv7. By means of experimental evaluation, we demonstrate that the proposed Residual YOLOv7 model significantly raises attack detection rates and reduces false positives when compared to conventional techniques. With 2.3% false positive rate, we showed a 97.5% threat detection accuracy in simulated IoT-SDN setups. The system's real-time processing powers ensure quick security measure implementation, which lowers the risk of prolonged exposure to threats. The adaptability of the response to various types of attacks also generally improves network resilience.