Detection of DDoS Attack Using Machine Learning Algorithms

Abstract

Abstract: Distributed Denial of Service (DDoS) attack is one of the common network attacks. DDoS attack occurs when a website or server is targeted by a malicious user to deny the services by flooding with unwanted information. This causes delay of services to legitimate user. Denial of Service (DoS) attack happens when the attack is from single source, whereas Distributed Denial of Service attack (DDoS) happens when the attack is from many number of sources say Botnet which controls the devices remotely for malicious purpose. A set of eight supervised machine learning algorithms are selected to detect DDoS attack and found the best model in terms of accuracy, precision, recal and false alarm ratel. For experimental results, a standard benchmark dataset CIC-IDS2017 is used for training and testing purpose. K-Fold cross validation is performed during the preprocessing stage. Then the eight models are trained and tested via K-Fold cross validation to find the best one to detect the DDoS attack at the earliest stage. In the testing phase we tested the trained models with the parameters Accuracy, Precision, Recall and FAR. Among eight models we found that Random Forest is the best model by considering all parameters into account. It has produced 99.885% accuracy, 99.88% Precision, 100% Recall and 0.05% False alarm rate to detect DDoS attack at the earliest.

Keywords: DDoS-Distributed Denial of Service, K-Fold Cross Validation, Machine Learning Algorithms