Analysis of IoT Botnets using Machine Learning Technique

Abstract

Abstract— Internet of Things (IoT) botnets have been used to bring down some of the biggest services on the Internet. The spread of Internet of Things (IoT) botnets like those utilizing the Mirai malware was successful enough to the most powerful DDoS attacks. Particularly, behavioral-based approaches suffer from the unavailability of the benchmark datasets and this lead to lack of precise results evaluation of botnet detection systems, comparison, and deployment which originates from the deficiency of adequate datasets. This project used machine learning as an algorithms program that learn to collect data. There are various data mining tools available to analyze data related IoT botnets detection. However, the problem arises in deciding the most appropriate machine learning techniques or algorithm on particular tools to be implemented on IoT botnet data. This research is focusing only on classification techniques. Hence, the objective of this research is to identify the best machine learning technique or algorithm on selected tool for IoT botnets detection. Five techniques: Random Forest, J48, JRip, Naïve Bayes and BayesNet. are selected and applied in selected tools namely Weka. The expected output of this project is to provide the machine learning techniques for effective detection of IoT botnets flows that have high predictive accuracy. This result provides an option for the researcher on applying technique or algorithm on selected tool when analyzing IoT botnets data.