Prevalent but Fragmented: Understanding the Scattered URL Navigation Interventions in Modern Web Clients

Modern Web clients (WebCs), software that can retrieve and display Web content, actively intervene in the navigation process of URLs when necessary—a behavior we refer to as navigation intervention. Typically, navigation interventions intercept malicious URLs via warning pages to protect security (e.g., via Google Safe Browsing). Nowadays, the platforms and types of WebC are increasingly diverse. In particular, in addition to browsers, feature-rich mobile applications with embedded WebViews (in-app WebCs) have become popular. The development of WebCs has expanded the scope of navigation interventions accordingly. In terms of explanation, WebCs have been performing content filtering. In terms of form, there are many other forms of interventions beyond warning pages. Meanwhile, differences in popular WebCs across regions—particularly between China and the rest of the world—also lead to variations in navigation interventions across different regions. The scattered development of navigation interventions in terms of their explanations, forms, and regions makes their current distribution and impact on users unclear. This study presents a large-scale measurement of navigation interventions across diverse WebCs. We developed WebCs URL Navigation Intervention Test Framework ( UNITF ), an automated cross-platform framework that combines Optical Character Recognition and Large Language Model technologies. Applying UNITF to 63 WebCs, we analyzed 1,806 (US) and 1,523 (CN) valid URLs derived from a 3,500-URL seed dataset (1,500 malicious and 2,000 non-malicious) after rigorous verification. Our study found that navigation interventions are widely used in WebCs (85.71%). However, there are differences in the distribution of intervention forms and explanations among different categories of WebCs. Moreover, there are significant differences between China’s navigation intervention practices and international practices. More importantly, we identified fragmentation in WebCs’ navigation interventions, including inconsistencies in functionality across WebCs and differences in regional performance within the same WebC, and provided the community with actionable recommendations and directions for improvement.