Performance Comparison of Machine Learning Models for DDoS Attacks Detection

Abstract

Distributed denial of service (DDoS) attack is one of the most costly attacks for IT system in terms of time and money. In this paper, the use of machine learning algorithms for DDoS detection has been addressed. The traditional SVM and new emerging deep learning algorithm, namely Deep Feed Forward (DFF), are evaluated. The DARPA Scalable Network Monitoring and DARPA 2009 DDoS attacks dataset is used to test the effectiveness of these two algorithms. The dataset is preprocessed to find the potential speedup of the classification process. From the experiments, DFF deep learning algorithm has achieved a high accuracy of 99.63% with the training time of 289.614 secs. For SVM, the highest accuracy achieved is 93.01%, with the training time of 371.118 secs. Anyway, SVM is able to deliver a faster classification time. Therefore, DFF is suitable for the situation when accuracy is the main concern while SVM can be used when speed of classification is a critical factor.