OWASP IoT Top 10 based Attack Dataset for Machine Learning

Abstract

Internet of Things (IoT) systems are highly susceptible to cyberattacks by nature with minimal security protections. Providing a massive attack surface for attackers, they automatically become easy targets with potentially catastrophic impacts. Researchers are currently focusing on developing various anomaly detection systems for IoT networks to deal with this situation. However, these systems require a comprehensive labeled attack dataset to classify the malicious traffic correctly. This paper presents a systematic approach to design and develop an IoT testbed to generate such an attack dataset, namely the AIoT-Sol Dataset. Our proposed dataset contains the benign traffic as well as the often-overlooked attack techniques in the existing IoT datasets. It encompasses 17 attack types from several categories: network attacks, web attacks, and web IoT message protocol attacks. We selected these attacks by referencing the Open Web Application Security Project (OWASP) IoT Top Ten. Also, we provide a mapping of possible attacks for all ten security risks.