CS&E Colloquium: Inferring Device Events and User Activities from IoT Network Traffic
The computer science colloquium takes place on Mondays from 11:15 a.m. - 12:15 p.m. This week's speaker, Guoliang Xue (Arizona State), will be giving a talk titled, "Inferring Device Events and User Activities from IoT Network Traffic".
Abstract
The availability of ubiquitous and heterogeneous Internet-of-Things (IoT) devices in smart homes and their interactions with users provide a unique opportunity to monitor, understand, recognize, learn, and infer device events and user activities for safety monitoring, connected health, as well as other disruptive services. In this talk, we discuss recent studies in inferring device events and user activities from IoT network traffic in a smart home environment. First, we study what kind of information can be extracted from mostly encrypted IoT network packets. It turns out that information in an IP packet such as timestamp, address of IoT device, server canonical name, remote port, protocol, packet length, etc., can be used to identify an IoT device event, with varying degrees of accuracy. Second, efficient algorithms are designed to infer a sequence of device events from the IoT network traffic log. Next, we explore the possibilities of identifying user activities from a sequence of IoT device events. Unsupervised learning algorithms are designed to infer a subset of user activities from the IoT network traffic log. These algorithms are validated using the network traffic of heterogeneous IoT devices collected at the router of a real-world smart home testbed.