Detection and Mitigation of DDOS Attack in SDN Using Feature Based RF & MLP Approach

Software Defined Networking (SDN) is an architecture that decouples the control plane from the data plane, allowing for centralized oversight and flexible configuration of network resources via software. However, this centralization of the controller makes it especially susceptible to Distributed Denial of Service (DDoS) attacks.. To effectively prevent and mitigate such attacks, deploying a robust machine learning model is crucial. In this project, the Genetic Algorithm and Chimp Optimization Algorithm were employed as optimal feature selection methods to identify and extract the most representative and distinguishing features from the SDN_DDoS_2020 dataset, which contains ICMP, TCP, and UDP attack traffic (sourced from Mendeley). The refined feature subset was then used to train and evaluate the performance of two machine learning algorithms commonly used are Random Forest (RF) and Multilayer Perceptron (MLP). Furthermore, a dedicated mitigation model was developed to block or redirect the flood requests from TCP, UDP, and ICMP attack traffic. Among the models tested, the combination of the Chimp Optimization Algorithm with the MLP classifier demonstrated superior performance, achieving an impressive accuracy of 98.68%. Notably, this model also exhibited a low false alarm rate of 0.74 compared to all other models, enhancing its reliability in real world scenarios where accurate attack detection and minimal false positives are essential.