Are You Comfortable Sharing It?: Leveraging Image Obfuscation Techniques to Enhance Sharing Privacy for Blind and Visually Impaired Users

Paper Title

Are You Comfortable Sharing It?: Leveraging Image Obfuscation Techniques to Enhance Sharing Privacy for Blind and Visually Impaired Users

Publication Info

• Topic area: Privacy-enhancing image-sharing techniques for Blind and Visually Impaired (BVI) users.
• Keywords: Blind and Visually Impaired (BVI), image obfuscation, privacy, sharing comfort, accessibility, audience-specific preferences, filtering techniques, contextual integrity, personalization, assistive technologies.

Background and Problem

• Problem / challenge: BVI individuals face privacy risks when sharing images, as they cannot visually verify content and may unintentionally disclose sensitive or inappropriate information. Existing tools lack empirical evidence on how obfuscation techniques affect BVI users’ comfort in sharing images across different audiences and content types.
• Significance: Addressing these privacy concerns is crucial to empower BVI users to confidently share images while maintaining their privacy and social relationships.
• Motivation and related work: Prior research explored privacy-preserving tools and obfuscation techniques but focused on usability or narrow contexts. Gaps remain in understanding how BVI users perceive different filtering methods across diverse sensitive content categories and how audience type influences their comfort in sharing.

Solution

• Proposed approach: A study evaluating BVI users’ preferences for four image obfuscation techniques (blurring, pixelation, masking, content-based filling) across ten sensitive content categories and three audience types (family, friends, strangers).
• Novelty:
  1. Systematic comparison of obfuscation techniques across diverse sensitive content categories.
  2. Analysis of how audience type affects BVI users’ comfort in sharing filtered and unfiltered images.
  3. Design guidelines for privacy-preserving image-sharing tools tailored to BVI users.
• Procedure and key techniques:
  • Participants rated comfort levels for sharing images before and after applying filters.
  • Four obfuscation techniques were evaluated: blurring, pixelation, masking, and content-based filling.
  • Ten sensitive content categories were tested, including nudity, identity information, private areas, and embarrassing shots.
  • Comfort ratings were analyzed for three audience types: family, friends, and strangers.

Results

• Concrete findings:
  • Blurring was preferred for highly sensitive categories (e.g., nudity, identity information).
  • Pixelation was the least preferred technique across all categories.
  • Filtering significantly increased comfort levels for sharing images, particularly with strangers.
  • No single obfuscation technique was universally preferred; preferences varied by content type.
• Advantage over baselines:
  • Empirical evidence on the relationship between content sensitivity, audience type, and obfuscation preferences, which was previously unexplored.
  • Demonstrated the importance of adaptive and context-aware filtering tools.
• Experiments / evaluation:
  • Conducted with 20 BVI participants (ages 20–31, 7 female).
  • Participants rated comfort levels using a 7-point Likert scale for sharing images with family, friends, and strangers.
  • Statistical analyses (Friedman and Wilcoxon tests) revealed significant differences in preferences and comfort levels across categories and audiences.
• Limitations and future work:
  • Limited diversity in image categories (10 categories, 2 images each).
  • Did not assess the utility of filtered images for practical tasks or receiver interpretation.
  • Participants were from a metropolitan city, limiting cultural and geographic diversity.
  • Future work should explore broader datasets, international contexts, and task-based evaluations.

Summary

This study investigates how BVI individuals evaluate image obfuscation techniques and how these affect their comfort in sharing sensitive images with family, friends, and strangers. Blurring was preferred for sensitive categories, while pixelation was consistently rated poorly. Filtering significantly increased comfort levels, particularly for sharing with strangers. The findings highlight the need for customizable, context-aware filtering tools that align with BVI users’ privacy expectations and social norms. Future work should expand on dataset diversity, cultural contexts, and task-based evaluations to refine these tools further.