Generic Application Layer Features For IoT Devices Identification

The wide adoption of Internet of Things (IoT) in traditional networks and critical infrastructures has brought many advantages. At the same time, insecure IoT devices provide a loophole in existing infrastructure that miscreants can exploit. Mirai incident is a well-known example, where attackers exploited the internet using IoT devices. Hence, a quick, accurate and energy-efficient IoT device identification mechanism is required to cope with these emerging challenges. In this research, we have proposed a generic set of application layer features for IoT device identification. To show the effectiveness of proposed generic application layer features we have compared them with network layer features using machine learning models and IoT devices traces. Previously most of the work was done on network layer features. However, for IoT devices network layer features show constancy in their pattern as compared to application layer features which are distinct and unique in nature. As a consequence, large number of network layer features are required for identification of devices causing use of higher prediction time, computational and feature extraction cost. We have also developed the first available open-source application layer feature extractor tool. Researchers can utilize this tool for acquiring application layer datasets and utilize them in different research domains.