Towards Labeling On-Demand IoT Traffic Daniel Campos TJ OConnor Florida Institute of Technology Florida Institute of Technology Melbourne, FL, USA Melbourne, FL, USA
[email protected] [email protected] ABSTRACT The lack of transparency and control in IoT devices has compli- A lack of transparency has accompanied the rapid proliferation of cated user security and privacy. A recent class-action lawsuit argued Internet of Things (IoT) devices. To this end, a growing body of work that Ring Inc. allowed attackers to access users’ cameras covertly exists to classify IoT device traffic to identify unexpected or sur- and surreptitiously [23]. In a separate incident, an ADT technician reptitious device activity. However, this work requires fine-grained pled guilty to remotely spying on customers over 9,600 times in a labeled datasets of device activity. This paper proposes a holistic four-year period [8]. Due to the limited interfaces on IoT devices, approach for IoT device traffic collection and automated event la- victims are left unaware of their devices’ actions or how to imple- beling. Our work paves the way for future research by thoroughly ment access controls [12]. IoT devices occupy our most personal examining different techniques for synthesizing and labeling on- spaces with cameras and microphones. Unfortunately, this lack of demand traffic from IoT sensors and actuators. To demonstrate this transparency and control has enabled intimate partner violence by approach, we instrumented a smart home environment consisting using IoT devices to spy and harass victims [2, 8, 9]. Labeling and of 57 IoT devices spanning cameras, doorbells, locks, alarm sys- identifying traffic flows by purpose (i.e., camera streaming) offers tems, lights, plugs, environmental sensors, and hubs.