Selection of efficient machine learning algorithm on Bot-IoT dataset for intrusion detection in internet of things networks

Abstract

With the growth of internet of things (IoT) systems, they have become the target of malicious third parties. In order to counter this issue, realistic investigation and protection countermeasures must be evolved. These countermeasures comprise network forensics and network intrusion detection systems. To this end, a well-organized and representative data set is a crucial element in training and validating the system's credibility. In spite of the existence of multiple networks, there is usually little information provided about the botnet scenarios used. This article provides the Bot-IoT dataset that embeds traces of both legitimate and simulated IoT networks as well as several types of the attacks. It provides also a realistic test environment to address the drawbacks of existing datasets, namely capturing complete network information, precise labeling, and a variety of recent and complex attacks. Finally, this work evaluates the confidence of the Bot-IoT dataset by utilizing a variety of machine learning and statistical methods. This work will provide a foundation to enable botnet identification on IoT-specific networks.