Seeing is Believing?: Effects of Visualization on Smart Device Privacy Perceptions

Carlos Bermejo Fernandez, Petteri Nurmi, Pan HuiACM Multimedia 2021 · DOI

TL;DR

This paper investigates how visualizing smart device data collection and location impacts user privacy perceptions. They found that AR visualizations, especially those showing exact location and data types, lead to more accurate privacy awareness than traditional interfaces, even for tech-savvy users. This research offers practical guidelines for designing more effective privacy assistants for the growing smart device ecosystem.

What Problem Does This Paper Address?

The proliferation of smart devices has created a significant privacy challenge. While users express concern about their data, they often fail to act on these concerns, a phenomenon known as the "privacy paradox." This discrepancy stems from a lack of understanding about what data is collected, how it's used, and the potential risks involved. Existing privacy assistants often rely on traditional, list-based interfaces that don't effectively contextualize this information, leaving users with a false sense of security. This paper aims to bridge this gap by exploring how different visualization techniques, particularly augmented reality (AR), can improve users' understanding of smart device privacy risks and lead to more accurate perceptions.

Methods

The researchers conducted two studies to explore the impact of visualization on privacy perceptions.

  • Study 1: Visualization Effects on Privacy Perceptions: This study used a mixed factorial design with smart device type (smart camera vs. smart speaker) and interface type (list-only, list with real-time data visualization, and list with real-time data visualization plus location) as independent variables. Participants interacted with these interfaces and then reported their comfort levels regarding privacy. The study also assessed the impact of presenting potential misuses of collected data.
  • Study 2: Crowdsourced AR Interface Evaluation: This study employed a within-subjects design using Amazon Mechanical Turk to evaluate user preferences for different AR interface designs. Participants were presented with various AR interface configurations (varying elements like video feed, icon-based menus, color-coding, and table layouts) and asked to identify collected vs. uncollected data, and then rate their preferred design. This study focused on smart cameras to ensure realistic visualization.

Datasets

Not discussed in the paper. The studies utilized simulated smart devices (Raspberry Pi with a webcam and speaker) and a OnePlus 3T device for the AR application.

Key Insights and Ideas

  • Seeing is Believing (and Worrying): Visualizing the exact location of a smart device and the specific types of data it collects significantly reduces user comfort levels, indicating a more accurate perception of privacy risks. This suggests that simply informing users isn't enough; they need to "see" the implications.
  • AR as a Privacy Enhancer: Augmented reality interfaces, by their nature, can provide a more intuitive and contextualized way to understand smart device privacy. This is particularly true when AR is used to overlay information directly onto the physical environment where the device is located.
  • Misuse Awareness is Key: Presenting users with examples of how their data could be misused, even in a benign way, dramatically lowers comfort levels across all interface types. This highlights that users' initial privacy perceptions can be overly optimistic due to a lack of awareness of potential threats.
  • Preference vs. Efficiency Trade-off: While participants preferred simpler AR interfaces (e.g., color-coded, icon-based) for their ease of use and speed, interfaces that provided more detailed information (like video feeds) led to more accurate privacy perceptions, even if they took longer to process.
  • Tech-Savvy Still Struggle: Even participants with experience in AR and IoT found it challenging to fully grasp the privacy implications of smart devices, underscoring the need for effective visualization tools.

Key Results

  • Study 1:
    • Visualizing device location significantly reduced comfort levels for both smart cameras (d=1.95) and smart speakers (d=0.92).
  • Presenting possible misuses of data led to a significant drop in general comfort (d=1.02).
  • AR interfaces generally resulted in lower comfort levels compared to traditional interfaces, suggesting they are more effective at revealing actual threats.
  • Study 2:
    • The fastest interface completion time was for a design without a video source (12.1 seconds).
  • However, participants preferred interfaces that were easier to read and provided clear visual cues, even if they took longer to complete tasks. A design with color-coding, icons, and a table layout was highly rated.
  • Interfaces with video feeds, while potentially more informative, had longer completion times and were not always the most preferred.

Limitations and Future Work

  • Ecological Validity: Study 1 was conducted in a controlled environment, which might not fully replicate real-world scenarios. The study also acknowledges the potential for carry-over effects between interface interactions.
  • Participant Demographics: While efforts were made to recruit a diverse group, the participants in Study 1 had some prior experience with AR, and the crowdsourced participants in Study 2 might not be fully representative of the general population.
  • AR Privacy Threats: The study did not address potential privacy threats associated with the use of AR itself.
  • Future Work:
    • Investigate more sophisticated AR interfaces, such as those integrated into smart glasses, and explore natural interaction methods like hand gestures.
  • Study privacy in semi-public and public locations, where sensor locations might be unknown.
  • Explore the integration of visualizations for microphone and camera coverage radius.
  • Further investigate how to effectively convey potential misuse of data through authoritative sources.
  • Develop AR-based privacy assistants that can transmit information between smart devices and PAs using location or Bluetooth for discovery.

Summary updated

BibTeX
@inproceedings{fernandez2021seeing,
  author = {Carlos Bermejo Fernandez and Petteri Nurmi and Pan Hui},
  title = {{Seeing is Believing?: Effects of Visualization on Smart Device Privacy Perceptions}},
  booktitle = {ACM Multimedia},
  year = {2021},
  doi = {10.1145/3474085.3475552},
}
Figure