Efficient DDoS Attack Detection using Machine Learning Techniques

Abstract

Distributed Denial-of-Service (DDoS) attacks are deliberate attempts to interrupt the regular traffic of a specific server, network, organization, by flooding the victim or its neighbouring servers with network traffic. Identification of such attacks using various models is challenging due to the substantial modifications in their regular pattern and traffic rates. An automated detection approach is used to mitigate this issue, by limiting the feature space, which minimizes the model's overfitting and computational time. The CICDDoS2019 data set containing extensive DDoS attacks are used to train and access the proposed methodology in a cloud-based context. The relevant features are extracted using the Extra Tree classifier and they are fed to the Decision Tree, XGBoost, and Random Forest. Consequently, the proposed model can be used to detect DDoS attacks effectively.