Real-time identification of device types based on a supervised classification

A wide diversity of IoT devices is being connected in various smart homes with a rapid growth. Identifying IoT devices being connected to the network ensure a better management and enable a security enforcement within the home network. This task is very challenging in such heterogeneous environment. In this paper, we present a classification approach based on network features extracted from network communication characteristics and packets payload to discriminate between devices traffic. Our solution automatically identify a new connected device to the home network. Furthermore, we evaluate the performance of our method on a representative and heterogeneous real IoT devices set. Our results show an average accuracy equal to 98% based a decision tree models using the one-vs.-all method.