PriView -- Exploring Visualisations Supporting Users' Privacy Awareness

Document Title

PriView – Exploring Visualisations to Support Users’ Privacy Awareness

Document Information

• Research Area: Privacy protection, Human-Computer Interaction, Internet of Things
• Keywords: Smart home, Smart environment, Smart devices, Internet of Things (IoT), Privacy protection, Thermal imaging camera, Mobile application, Augmented Reality (AR), Virtual Reality (VR), Visualization

Research Background and Issues

• Problems or Challenges:

  • As the number of sensors in environments increases, users often lack awareness of the location and behavior of privacy-invasive devices (e.g., cameras, microphones).
  • Existing privacy notifications (e.g., privacy policies) are often unreadable and not user-friendly, making it difficult for most users to extract key information.
  • Privacy risks posed by devices such as smart speakers and cameras are not intuitive in public and private environments, requiring users to actively investigate to ensure privacy safety.

• Importance:

  • Enhancing users' privacy awareness can help them avoid certain risk areas or choose not to disclose sensitive information, thereby better protecting privacy in daily life.
  • Privacy-focused design can assist users in managing their private data more effectively in increasingly complex smart environments.

• Research Motivation and Related Work:

  • Related work has proposed privacy labels, privacy visualization tools, and privacy notification designs, but there is still room for improvement in user perception and usability.
  • The opacity of smart devices (e.g., whether the device is currently recording data) makes it difficult for users to understand the flow of their private data, exacerbating privacy risks.

Solution

• Methods and Solutions:

  • A privacy visualization concept—PriView—is proposed, enabling users to intuitively perceive the privacy risks of surrounding smart devices through visualization.
  • Two prototypes were developed: a mobile application based on thermal imaging cameras and VR-simulated scenarios (including rental apartments, public transportation stations, etc.).

• Innovations:

  • Using thermal imaging cameras to detect device activity and visually presenting potential privacy risks through intuitive visualizations (e.g., borders, text labels, 3D shapes).
  • Combining user experiments with VR scenario simulations to study users' needs and preferences for privacy visualization across multiple environments and use cases.

• Implementation Steps and Techniques:

  • Mobile application: Detect device thermal states using the FLIR One thermal imaging camera and identify smart devices and their on/off status using the YOLO deep learning model.
  • VR model: Create six sample scenarios (e.g., rental apartments, office kitchens) developed using the Unity game engine and displayed on HTC Vive Pro headsets to showcase various visualization options.
  • User experiments: Compare user preferences, usability, and comprehension across two output devices and seven visualization representations.

Research Outcomes

• Specific Findings:

  • Participants generally endorsed the concept and design of PriView, recognizing its potential not only to enhance privacy awareness but also to support other applications such as device maintenance and removal.
  • Users preferred detailed visualizations in unfamiliar environments (e.g., rental apartments) but favored simpler prompts in familiar and trusted environments (e.g., friends' homes).
  • "Text labels" and "3D shapes" were considered the most effective ways to provide information, while "warning icons" were deemed insufficient in conveying adequate information.

• Advantages of Existing Solutions:

  • Compared to traditional privacy notifications, PriView provides privacy-related information in a more intuitive and user-friendly manner.
  • Dynamically detecting device states and displaying privacy-related areas within scenarios makes it suitable for complex and dynamic smart environments.

• Experiment or Evaluation Results:

  • User experiments revealed that segmenting scenarios and tailoring solutions to user needs can achieve better user experiences.
  • The mobile application received a usability score of 71.14 (SUS), while the HMD experiment scored 73.85, indicating high acceptability for both.
  • Most users preferred "text labels" and "3D shapes" when using head-mounted displays.

• Limitations and Future Directions:

  • Limitations: The experimental environment was limited, failing to cover all scenarios that might affect device detection or privacy visualization effectiveness; the sample was biased toward young student groups.
  • Future Directions:
    • Enhance the generalizability of device detection algorithms, such as through large-scale training datasets or collaboration with device manufacturers to access relevant information.
    • Design flexible and personalized interfaces to meet users' multi-layered needs for detail, visual information, and interaction.
    • Explore how to balance privacy and data openness in public and private settings.

This study provides a new perspective on privacy protection in smart environments and validates its effectiveness and potential through user experiments. Further optimization research can improve its applicability and scalability.