Detection and mitigation of DDoS attacks in SDN networks

Abstract

Software-De ned Networking (SDN) is an emerging concept designed to substitute traditional networking by breaking up vertical integration. Central control is the biggest bene t of SDN, but a single point of failure is also a failure if a distributed denial of service (DDoS) attack makes it unattainable. This memory provides an e cient solution based on machine learning (ML) algorithms to detect and mitigate DDoS attacks with the help of Mininet and the Ryu controller to simulate the network. In contrast, DDoS attacks were simulated using of Hping3 tool. This memory proves that the type of topology is signi cant and can a ect the percentage of success of DDoS attacks in such networks. Six supervised ML algorithms (LR, K-NN, NB, SVM, DT, and RF) were tested and evaluated using a synthetic dataset. The results show that DT and RF are the best compared to the other algorithms with 100% of accuracy. The proposed system shows its e ciency in detecting and mitigating DDoS attacks with the RF classi er and only ve features. At the same time, the mitigation was provided by adding a ow rule to the switch to drop the malicious tra c. Keywords: Software-De ned Networking (SDN), Distributed Denial of Service Attack (DDoS), Machine learning (ML).