Enhancing Network Resilience for Flood Response and Rehabilitation Using SDN

Among the most costly and deadly natural calamities are floods. Of all the facilities, electricity and communication networks are the most vulnerable to flooding and are affected first. Floods pose a serious threat to the physical infrastructure of communication and electrical networks, which is essential for reaction and recovery operations. This research uses Software-Defined Networking (SDN), machine learning, and image processing to present a novel method for flood-based natural catastrophe prediction, response, and rehabilitation. The study investigates the use of unsupervised KMeans clustering to distribute nodes optimally across satellite pictures and the application of the supervised CatBoost algorithm combined with log transformation to predict areas impacted by flooding. The SDN controller considered in the simulation is the Open Network Operating System with inherent self-healing, next-generation device support, and high performance. The virtual network setup for hosts, switches, and links on ONOS is established using Mininet. The entire SDN setup is deployed using a Docker container to ensure scalability. The simulation shows how a strong mesh network set up with SDN may self-heal and improve network performance.