IoT Event Classification Based on Network Traffic

Abstract

The Internet of Things (IoT) consists of sensors and actuators that facilitate many aspects of our daily life. Compared to typical computing devices such as laptops and smartphones, these devices have a very limited set of functionalities and states. Researchers have shown that it is possible to infer the device type from its network traffic. In this paper, we show that an external observer that sniffs the network traffic of an IoT device can further classify device events and hence infer user actions by employing machine learning classifiers. We evaluate and compare the performance of ten machine learning algorithms in classifying 128 device events from 39 different devices. We analyze the impact of the user interaction through LAN and WAN as well as controllers such as Alexa voice assistant on the correct classification of device actions. We also inspect whether the region from which the device is impacts the performance of classifiers as researchers have shown that differing privacy restrictions lead to different external communications.