The rapid proliferation of Internet of Things (IoT) technology in various sectors has created new security and management challenges. The cornerstone of any effective security strategy is understanding what assets need protection and discovering the devices connected to the network. The correct identification of an IoT device plays a crucial role in IoT security solutions \cite{PaloAltoNetworks2023, checkpoint, firedome}. Extensive research carried out in both the industry and academia has been dedicated to IoT identification algorithms \cite{le2019policy,guo2018ip,hafeez2020iot,marchal2019audi,aneja2018iot}. These algorithms rely primarily on labeled IoT datasets for the real-time identification of previously \textbf{seen} devices. The algorithms learn each device’s unique network properties and then identify them. However, in the IoT realm, maintaining a lab with every possible IoT device is simply not feasible.
This poster argues that IoT labeling is an imperative challenge and introduces an approach that \textit{labels} unseen IoT devices by their type: \textit{vendor, function}.
The motivation to address it was born from discussions with IoT security companies that have difficulty labeling IoT devices in the wild.