Efficient Feature Selection for Detecting Botnets based on Network Traffic and Behavior Analysis

Ensuring integrity and security of computer networks is one of the growing concerns. The number of malware specifically designed to damage, disrupt or perform illegitimate actions on data, networks or hosts are increasing day by day. Detection of hosts infected by malware known as bots is the main focus of this paper. While Botnets are an emerging threat with hundreds of millions of computers infected, the research and solutions of it are still in their infancy stage. In this paper, at first we propose a feature selection algorithm to reduce extracted features from network flows. The selected features are lately analyzed using a supervised machine learning technique to effectively detect the presence of botnets. The experimental evaluation based on a versatile existing data set shows that the proposed model is able to effectively detect bots with more accuracy and high detection rate with moderate false alarms in the botnet’s Command and Control (C&C) phase.