Machine Learning based Botnet Detection in Large-Scale Network

Botnets is a term given to a network of computational devices which are affected with malware which is used to bring them under the control of an attacker. These computers enslaved by the attacker are used for various attacks such as DoS attack, phishing attack, stealing of data, Crypto-Jacking etc. The attacker does so with the help of a command-and-control server. This command-and-control server gives the necessary commands to bots which are the enslaved computational devices. The Internet of Things has become an undeniable part of today’s computing world. Recently IoT devices are extensively affected by the botnets as they have low security compared to the normal computing devices. Hence, the security of IoT devices must be ensured. A study of the recent works in the field leads us to the conclusion that most of the works are specified to the IoT devices only. Today’s internet, the normal computational devices as well as the IoT devices are becoming so much connected. So along with the IoT the security of normal computational devices must also be addressed. The network traffic is becoming so large. The classification of such a large network will demand a large processing power which may not be affordable by all. So, a more affordable approach in this field must be devised. This paper proposes a Hadoop Distributed File System based system, which incorporates the power of multiple systems for the classification of network traffic in a very large network. This study considers the recent works in the field. On the insight of the recent works, the system uses CTU-13 dataset for training the model. The classification algorithm used is the Random Forest algorithm.