Understanding factors behind IoT privacy -- A user's perspective on RF sensors

TL;DR

This study challenges the prevailing view that RF sensing is inherently privacy-preserving by showing that non-expert users weigh not only the data collected but also the inferences that can be drawn, along with device familiarity and user control. Through an online survey (162 US participants) comparing cameras, mmWave RF sensors, and WiFi routers, the authors demonstrate that data-inference combinations can variably influence privacy perceptions across modalities; for non-human-interpretable data, inferences drive perceived privacy more strongly, while for human-interpretable data, data and inferences exert comparable influence. The findings reveal that device appearance and familiarity shape privacy judgments, and that granting users comprehensive control over device design and data policies significantly increases willingness to deploy sensing devices in private spaces. Collectively, the work calls for transparent disclosure of inferred capabilities and stronger user-centric controls to align IoT deployments with user privacy expectations in the ML era.

Abstract

While IoT sensors in physical spaces have provided utility and comfort in our lives, their instrumentation in private and personal spaces has led to growing concerns regarding privacy. The existing notion behind IoT privacy is that the sensors whose data can easily be understood and interpreted by humans (such as cameras) are more privacy-invasive than sensors that are not human-understandable, such as RF (radio-frequency) sensors. However, given recent advancements in machine learning, we can not only make sensitive inferences on RF data but also translate between modalities. Thus, the existing notions of privacy for IoT sensors need to be revisited. In this paper, our goal is to understand what factors affect the privacy notions of a non-expert user (someone who is not well-versed in privacy concepts). To this regard, we conduct an online study of 162 participants from the USA to find out what factors affect the privacy perception of a user regarding an RF-based device or a sensor. Our findings show that a user's perception of privacy not only depends upon the data collected by the sensor but also on the inferences that can be made on that data, familiarity with the device and its form factor as well as the control a user has over the device design and its data policies. When the data collected by the sensor is not human-interpretable, it is the inferences that can be made on the data and not the data itself that users care about when making informed decisions regarding device privacy.

Figures (8)

• Figure 1: We use 3 main classes of devices in our study. Each class has 2 different devices (in order to ascertain that the user responses are dependent on the class of device and not on the device itself.) These devices are a) An Amazon Blink home camera, b) Kasa Inoor Pan/Tilt Smart Security Camera c) A Wayv mmWave radar from Aienstein AI, d) LifeSmart mmWave Human Presence Sensor, e) LinkSys MAX-STREAM AC1300 WiFi router, and f) TP-Link AX1800 WiFi 6 router. In this paper, we use RF devices to learn more about factors that govern user privacy perceptions.
• Figure 2: User comfort level when shown (a) an image of the camera, (b) an image of the camera and a snapshot of the data that it collects, (c) an image of the camera and list of inferences that can be made on the data that it collects (assuming that only inferences and not data are being shared)
• Figure 3: User comfort level when shown (a) an image of the mmWave radar, (b) an image of the mmWave radar and a snapshot of the data that it collects, (c) an image of the mmWave radar and list of inferences that can be made on the data that it collects (assuming that only inferences, not data, are being shared)
• Figure 4: User comfort level when shown (a) an image of the WiFi router, (b) an image of the WiFi router and a snapshot of the data that it collects, (c) an image of the WiFi router and list of inferences that can be made on the data that it collects (assuming that only inferences and not data are being shared)
• Figure 5: User comfort level when shown (a) an image of a generic WiFi router, (b) an image of a WiFi router that looks like a microphone, (c) an image of a WiFi router that looks like a camera.