Improved Botnet Attack Detection Using Principal Component Analysis and Ensemble Voting Algorithm

Abstract

In recent times especially in the field of cloud computing, one of the most radical forms and threatening key issue of cyber-attacks has to do with botnets. Botnets with their flexible and dynamic nature together with a botmaster, mastermind their operations, change their codes, and update the bots daily in order to prevent the present detection methods. Despite high-profile efforts to tackle botnets, the number of botnets and infected systems only continues to grow. Early detection and analysis of these increasing number of botnet attack greatly impact the operational activities of any internet-related organization. Machine learning algorithms have played a key role in the detections and analysis of botnet infected packets in attacks such as DDoS attacks. This study, using Principal Component Analysis and an ensemble voting classifier improves the detection of botnet attacks. The results showed increased performance in terms of running time, accuracy, precision, and false-positive.