An Effiective IoT Device Identification Using Machine Learning Algorithm

Abstract

With rapid growth of the number of IoT device, there are more and more challenges in the secure manage for numbers of vulnerable IoT devices in practical network environment. One effective solution to this challenge is to develop a smart system which can identify the type of a device quickly and precisely. To aim this purpose, an advanced device identification method is presented in this paper. First, features during periodic flow inference and protocol inference are extracted to form the device fingerprints, and then a machine learning based classifier is used to identify the device type by using the importance of features. Experiment results show that not only the known types within a SOHO network such as smart speakers, cameras and sweeping robots can be identified successfully with an accuracy of 95%, but also new types can be classified without labeled data.