Detection and Analysis of DDoS Attacks in Internet Kiosk Voting Using Machine Learning Algorithms

Abstract

Classification algorithms recognize and differentiate class instances in a dataset to produce correctly classified output for better understanding of Distributed Denial-of-Service (DDoS) flood attacks captured on a workstation thereby improving availability of the e-voting server. Machine learning algorithms have been applied as detection mechanisms on DDoS attacks in securing network infrastructure by training the algorithms using datasets containing captured DDoS flood traffic on the network. In this paper, we compare and analyze the performance of Random Forest, Naïve Bayes and Multilayer Perceptron (MLP) machine learning classification algorithms on a sample of the Knowledge Discovery and Data Mining (KDD) Cup 99 Dataset containing four classes of DDoS attack using accuracy, precision and recall performance metrics. The training and testing of these classifiers on the dataset records was carried out in Waikato Environment for Knowledge Analysis 3.8.2 version tool using nine (9) best optimal attributes selected to produce confusion matrices at a reduced building model time. An accuracy of 98.65% in classifying DDoS flood attacks was achieved by MLP classifier. The study showed that the MLP classifier provides a better mechanism for DDoS detection for secure Internet voting system by increasing voting server’s performance in terms of system’s availability to voters during election process.