BAIT: Visual-illusion-inspired Privacy Preservation for Mobile Data Visualization

TL;DR

BAIT addresses privacy risks from shoulder surfing in mobile data visualizations by overlaying a decoy visualization designed through HVS-informed channels. It introduces a perception-driven optimization framework that balances proximity readability with far-distance privacy via distance-dependent visual alterations and a two-stage decoy generation process. The approach is validated through two user studies, demonstrating that BAIT preserves readability for legitimate users at close range while reducing shoulder-surfing accuracy to near-random levels at a distance, outperforming baseline masking methods. The work contributes a fully automated decoy-based privacy mechanism, design guidelines for illusion-based privacy in visualizations, and evidence for practical deployment without hardware modifications. It also outlines limitations and avenues for extending BAIT to more visualization types and animated content, highlighting its potential for real-world privacy protection on mobile devices.

Abstract

With the prevalence of mobile data visualizations, there have been growing concerns about their privacy risks, especially shoulder surfing attacks. Inspired by prior research on visual illusion, we propose BAIT, a novel approach to automatically generate privacy-preserving visualizations by stacking a decoy visualization over a given visualization. It allows visualization owners at proximity to clearly discern the original visualization and makes shoulder surfers at a distance be misled by the decoy visualization, by adjusting different visual channels of a decoy visualization (e.g., shape, position, tilt, size, color and spatial frequency). We explicitly model human perception effect at different viewing distances to optimize the decoy visualization design. Privacy-preserving examples and two in-depth user studies demonstrate the effectiveness of BAIT in both controlled lab study and real-world scenarios.

Figures (13)

• Figure 1: Illustration of our method. A user can discern the original visualization content from the privacy-preserving visualization at a close viewing distance. At a far distance, a shoulder surfer perceives only the decoy visualization from the privacy-preserving visualization.
• Figure 2: The contrast sensitivity function curve illustrates the relationship between spatial frequency, stimulus contrast, and human visual perception. (A) highlights the association between luminance contrast and spatial frequency, whereas (B) depicts the correlation between chromatic contrast and spatial frequency.
• Figure 3: Method Overview: Input: Original visualization. Visual variables in our privacy-preserving visualization. Visualization-dependent channels. Visualization-agnostic channels. A set of decoy visualizations generated based on the visual variables in both and . A set of processed original visualizations generated based on the varying values of a channel: spatial frequency in . A collection of privacy-preserving visualizations created by stacking the set of decoy visualizations and the set of processed original visualizations together. A perception-driven model that efficiently identifies and selects the most effective privacy-preserving visualization from the collection, ensuring optimal balance between visibility and privacy.
• Figure 4: Illustrations of decoy visualization generation for four popular visualization types. Bar chart: Manipulating the number and height of bars to obscure the original visualization; Line chart: Modifying the decoy lines' position, length, and tilt; Scatter plot: Changing the decoy circles' position and number; Pie chart: Altering the angles of decoy slices and enlarging the decoy pie chart.
• Figure 5: Examples of privacy-preserving visualizations produced by BAIT. The figure showcases the simulated perception of original visualizations at a close distance (top row), privacy-preserving visualizations at a close distance (second row), privacy-preserving visualizations at a far distance (third row), and decoy visualizations at a far distance (bottom row). Note: the effect of privacy-preserving visualizations (second row) depends on their displayed sizes, and readers are recommended to expand them to validate their effectiveness if they are displayed in a small size.