Authors:
Alejandro Guerra-Manzanares
;
Jorge Medina-Galindo
;
Hayretdin Bahsi
and
Sven Nõmm
Affiliation:
Department of Software Science, Tallinn University of Technology, Tallinn, Estonia
Keyword(s):
Botnet, Internet of Things, Dataset, Intrusion Detection, Anomaly Detection, IoT.
Abstract:
The exponential growth of the Internet of Things in conjunction with the traditional lack of security mechanisms and resource constraints associated with these devices have posed new risks and challenges to security in networks. IoT devices are compromised and used as amplification platforms by cyber-attackers, such as DDoS attacks. Machine learning-based intrusion detection systems aim to overcome network security limitations relying heavily on data quantity and quality. In the case of IoT networks these data are scarce and limited to small-sized networks. This research addresses this issue by providing a labelled behavioral IoT data set, which includes normal and actual botnet malicious network traffic, in a medium-sized IoT network infrastructure (83 IoT devices). Three prominent botnet malware are deployed and data from botnet infection, propagation and communication with C&C stages are collected (Mirai, BashLite and Torii). Binary and multi-class machine learning classification
models are run on the acquired data demonstrating the suitability and reliability of the generated data set for machine learning-based botnet detection IDS testing, design and deployment. The generated IoT behavioral data set is released publicly available as MedBIoT data set∗.
(More)