Software-Defined Networking (SDN) is a networking paradigm that has redefined the term network by making the network devices programmable. SDN helps network engineers to monitor the network expeditely, control the network from a central point, identify malicious traffic and link failure in easy and efficient manner. Besides such flexibility provided by SDN, it is also vulnerable to attacks such as DDoS which can halt the complete network. To mitigate this attack, the paper proposes to classify the benign traffic from DDoS attack traffic by using machine learning technique. The major contribution of this paper is identification of novel features for DDoS attack detections. Novel features are logged into CSV file to create the dataset and machine learning algorithms are trained on the created SDN dataset. Various work which has already been done for DDoS attack detection either used a non-SDN dataset or the research data is not made public. A novel hybrid machine learning model is utilized to perform the classification. Results show that the hybrid model of Support Vector classifier with Random Forest (SVC-RF) classifies the traffic with the highest testing accuracy of 98.8% with a very low false alarm rate.

软件定义网络（SDN）是一种网络范例，通过使网络设备可编程来重新定义术语网络。SDN可以帮助网络工程师快速，有效地监视网络，从中心位置控制网络，识别恶意流量并链接失败。除了SDN提供的这种灵活性外，它还容易受到DDoS等攻击的攻击，这些攻击可能会导致整个网络中断。为了减轻这种攻击，本文提出通过使用机器学习技术从DDoS攻击流量中区分良性流量。本文的主要贡献是识别DDoS攻击检测的新颖功能。将新功能登录到CSV文件中以创建数据集，并在创建的SDN数据集上训练机器学习算法。使用非SDN数据集或未公开研究数据的DDoS攻击检测已完成的各种工作。一种新颖的混合机器学习模型被用来执行分类。结果表明，支持向量分类器与随机森林的混合模型（SVC-RF）可以对流量进行分类，具有最高98.8％的测试准确度和极低的误报率。