Date of Award

5-2021

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Engineering and Sciences

First Advisor

TJ O’Connor

Second Advisor

Meredith Carroll

Third Advisor

Siddhartha Bhattacharyya

Fourth Advisor

Philip Bernhard

Abstract

A lack of transparency has accompanied the rapid proliferation of Internet of Things (IoT) devices. To this end, a growing body of work exists to classify IoT device traffic to identify unexpected or surreptitious device activity. However, this work requires fine-grained labeled datasets of device activity. This paper proposes a holistic approach for IoT device traffic collection and automated event labeling. Our work paves the way for future research by thoroughly examining different techniques for synthesizing and labeling on-demand traffic from IoT sensors and actuators. To demonstrate this approach, we instrumented a smart home environment consisting of 57 IoT devices spanning cameras, doorbells, locks, alarm systems, lights, plugs, environmental sensors, and hub. We release a sample dataset consisting of 16,576 labeled events over 467,883 network flows. Our results indicate that vendor APIs, trigger-action frameworks, and companion notifications can be used to generate scientifically valuable labeled datasets of IoT traffic and can used to automatically produce future datasets.

Download

Share

COinS