Analysis and detection of IoT botnets using machine learning

Botnets have become a common tool for cybercriminals to launch cyberattacks against organisations and individuals for monetary gain from zombie computers controlled by a central command and control system. Cyber security systems must be able to analyse and detect botnets to prevent harm to information systems. Traditional methods to analyse and detect botnet behaviour and attacks are not well suited to IoT devices, often too heavy on resources on a resource-limited IoT device. Further, detection has been made a critical issue of cyber security as cybercriminals are now targeting the growing and security vulnerable IoT infrastructure. Using pre-trained machine learning architectures as an alternative for botnet detection to overcome security shortcomings on IoT devices has become a viable option. In this review of the current research literature about the analysis and detection of IoT botnets using machine learning, several options are put forward with different datasets, feature selection techniques, machine learning algorithms and post-result processing methods to produce high accuracy, precision and recall rates for the detection of IoT botnets. From reviewing these different approaches, specific feature selection techniques and machine learning algorithms stand out as promising candidates for future research and applications in real-world environments for IoT botnet detection.