Automatic Recognition of Identification Schemes for IoT Identifiers based on A Hybrid DNN Model

Xiaotao Li, Shujuan You, Wai Chen

Abstract: In Internet of Things (IoT), a unique identifier is essential for each object to serve as its digital identity and helps users to obtain the detailed information about this object. However, there are multiple encoding rules being used by different organizations to identify their objects, thus creating a great challenge for the applications that need to resolve multi-source heterogeneous identifiers. To address this issue, we propose an algorithm to automatically recognize the identification schemes of heterogeneous IoT identifiers, based on a hybrid deep neural network (DNN) model that combines a one-dimensional DNN model and a two-dimensional convolutional neural network (CNN) model. To satisfy the input criteria of the two-dimensional CNN, a novel identifier-to-polygon (I2P) strategy is presented to transform an IoT identifier to a unique polygon image. We evaluate our algorithm on the ID-20 dataset, showing that we can achieve an IoT identifier recognition accuracy of up to 94.77% and outperform other state-of-the-art methods.