PriviAware: Exploring Data Visualization and Dynamic Privacy Control Support for Data Collection in Mobile Sensing Research

Document Title

PriviAware: Exploring Data Visualization and Dynamic Privacy Control Support for Data Collection in Mobile Sensing Research

Document Information

• Subject Area: Human-Computer Interaction, Privacy Control, Mobile Sensing Data Collection
• Keywords: Mobile Sensing, Data Collection, Privacy Control, Usable Privacy Design, Privacy Awareness Enhancement

Research Background and Problem Statement

• Problems and Challenges:

  • Large-scale, passive mobile sensing data collection is increasingly significant in advancing digital health research, but privacy issues and data transparency have become critical challenges. For example, passive collection of health-related information can raise ethical and privacy concerns.
  • Participants often lack awareness of the types of sensed data and their potential privacy risks, leading to an underestimation of privacy threats.
  • Existing privacy mechanisms are often static and lack user engagement.

• Why It Matters:

  • Behavioral and mental health research involves sensitive data, such as users' geographic locations and communication records, where data breaches could impact employment, insurance, etc.
  • Enhancing transparency and control over privacy threats is a necessary path to support research while protecting user privacy.

• Related Work and Research Motivation:

  • Existing studies mostly focus on single data types (e.g., GPS) or privacy management for mobile app permissions, with insufficient research on comprehensive and flexible privacy support for mobile sensing data collection.
  • The authors aim to improve traditional static privacy mechanisms by exploring the effects of data visualization and contextual filtering in enhancing user privacy awareness.

Solution

• Proposed Method:

  • Designed and developed a mobile application called "PriviAware," integrating data exploration and contextual filtering features.
  • Data exploration makes data collection details transparent through visualization.
  • Contextual filtering allows users to tailor data collection based on specific times and locations, enhancing fine-grained privacy control.

• Innovations:

  • Implemented a dynamic privacy control mechanism, enabling users to selectively disable data collection in specific contexts in real time.
  • Enhanced users' perception of privacy threats through user-friendly privacy visualization.

• Implementation Steps and Technical Highlights:

  1. Data Exploration Feature: Designed appropriate graphical representations (e.g., stacked bar charts, maps) for different data types (e.g., categorical data, geographic data).
  2. Contextual Filtering Feature: Supported pausing data collection based on time ranges or geographic locations.
  3. Conducted a three-week user experiment, dividing participants into two groups: one using only data exploration (Group A) and the other using both data exploration and contextual filtering (Group B).
  4. Evaluated participants' privacy awareness and behavioral changes through surveys and follow-up interviews.

Research Outcomes

• Specific Findings:

  • The data exploration feature effectively increased users' privacy awareness, enabling them to more clearly understand the collection and use of sensed data.
  • The contextual filtering feature enhanced users' sense of control over their privacy, significantly alleviating concerns in privacy-sensitive contexts.
  • User feedback indicated that the application interface was intuitive and met their privacy needs.

• Advantages Over Existing Solutions:

  • Outperformed traditional privacy notification mechanisms (e.g., static notifications and one-by-one permission controls) by providing more flexible privacy support.
  • The combination of data exploration and contextual filtering created a synergistic effect, helping users understand the details of data collection while enabling dynamic privacy configurations.

• Experimental Results:

  • Privacy awareness significantly increased among participants in both groups, especially for sensitive data types (e.g., GPS, app usage records).
  • Participants in Group B demonstrated higher engagement and proactivity through contextual filtering.
  • Although survey comparisons showed no significant inter-group differences, interview results highlighted the additional psychological value of contextual filtering.

• Limitations and Future Directions:

  • Limitations:
    • Experiment participants were primarily university students, leading to insufficient sample diversity.
    • The three-week experiment duration was relatively short, preventing exploration of long-term usage effects.
    • Contextual filtering was limited to time and location, without considering other factors (e.g., data recipients, data purposes).
  • Future Directions:
    • Expand to more diverse populations and more complex contexts (e.g., multiple contextual factors).
    • Explore data-driven automated filtering mechanisms to reduce users' decision-making burden.
    • Deploy the PriviAware system long-term to study the sustainability of privacy awareness changes.

Conclusion

By combining data visualization and dynamic contextual filtering, PriviAware achieves a more user-centered privacy control mechanism, significantly enhancing participants' privacy awareness and proactive behaviors. This research provides new perspectives in the field of privacy design and offers valuable insights for addressing privacy issues and user authorization management in mobile sensing data collection.