Equitable Machine Learning Algorithms to Probe Over P2P Botnets

Abstract

Cyber security has become very significant research area in line due to the increase in the number of malicious attacks by both state and nonstate actors. Ideally, one would like to properly secure the machines from being infected by viruses of any form. Nowadays, botnets have become an integral part of the Internet and the main drive for creating them is for financial gain. A bot conceals itself using a secret canal to communicate with its governing command-and-control server. Botnets are well-ordered from end to end using protocols such as IRC, HTTP, and P2P. Of all HTTP-based and IRC-based, P2P botnet detection became a challenging task because of its decentralized nature. The paper focuses on the techniques that are predominantly used in botnet detection and we formulate a method for detecting the P2P botnets using supervised machine learning algorithms such as random forest (RF), multilayer perceptron (MLP), and K-nearest neighbor classifier (KNN). We analyze the performance of selected algorithms there by revealing the best classification algorithm for detecting P2P botnets.