BAGEL: An Approach to Automatically Detect Navigation-Based Web Accessibility Barriers for Keyboard Users

Document Title

BAGEL: An Approach to Automatically Detect Navigation-Based Web Accessibility Barriers for Keyboard Users

Document Information

• Subject Area: Web Accessibility Technology/Software Testing
• Keywords: Web Accessibility, WCAG, Keyboard Navigation, Software Testing, UI Accessibility

Research Background and Problem Statement

• Identified Problems or Challenges:

  1. Keyboard navigation is crucial for many users with disabilities, yet many modern websites fail to meet Web Content Accessibility Guidelines (WCAG) in terms of keyboard usability.
  2. Dynamic updates via JavaScript may not be properly rendered for keyboard users, leading to navigation failures (KNF).
  3. Existing static analysis techniques are insufficient to uncover navigation issues.

• Significance: Design flaws in websites hinder effective navigation for keyboard users, affecting information access and service usage for approximately 15% of the global population with disabilities.

• Research Motivation and Related Work: Nearly 51% of users with disabilities rely on the web as a lifeline, yet fewer than 4% of top websites comply with the widely adopted WCAG standards. This study aims to address accessibility deficiencies related to keyboard navigation and provide higher precision and recall compared to existing detection tools.

Solution

• Proposed Method: A method named BAGEL is proposed to automatically detect navigation issues using dynamic web crawling techniques to identify keyboard navigation failures (KNF).

• Innovations:

  1. The first formal definition and modeling of keyboard navigation failure (KNF);
  2. Fully automated KNF detection technology surpassing the performance of existing tools;
  3. Analysis of UI navigation behavior through visual semantic models.

• Implementation Steps and Key Techniques:

  1. KFFG Modeling:
     • Create a keyboard navigation model to represent the focus navigation flow of keyboard users in the UI;
     • Dynamically crawl web pages and simulate keyboard operations (e.g., Tab switching).
  2. Semantic Clustering:
     • Use DBSCAN density clustering to group page elements into similar functional sets.
  3. Visual Clustering Analysis:
     • Compare screenshots to analyze visual differences during focus changes, detecting whether focus indicators are apparent.
  4. KNF Detection:
     • Analyze navigation order for intuitiveness using the model;
     • Detect unexpected focus changes during page transitions;
     • Identify insufficient visual focus issues.

Research Outcomes

• Specific Results:

  1. The BAGEL tool successfully achieved automatic detection of keyboard navigation barriers, with an average F1 score of 90%;
  2. Experiments conducted on 20 real-world web pages demonstrated BAGEL's high precision and recall in detecting KNFs.

• Advantages Over Existing Solutions: Compared to existing accessibility scanning tools (e.g., WAVE and QualWeb), BAGEL provides more detailed detection of navigation barriers, avoiding the high false positive rates and low recall of traditional static analysis tools.

• Experimental or Evaluation Results:

  • BAGEL detected more KNFs than other tools with lower error rates;
  • The tool's average runtime was approximately 8 minutes, slightly slower than existing tools but with higher accuracy;
  • Detected KNFs included unordered navigation, unexpected page transitions, and invisible keyboard focus.

• Limitations and Future Directions:

  • Limitations: Only addresses the initial state of a single page and cannot analyze dynamically loaded content.
  • Future Directions: Integrate a more comprehensive crawling mechanism to explore deeper page content and ensure compatibility with other browsers to detect cross-browser accessibility issues.