Privacy policies are the fundamental approach for conveying privacy notices to mobile app users. However, privacy policies are often lengthy and difficult for users to read and understand. To improve readability and engagement, researchers have proposed contextual privacy policies (CPPs) to display concise policy snippets only in the relevant contexts of the app’s graphical user interface (GUI). In this work, we formulate CPPs in mobile application scenario and present SeePrivacy, a multimodal framework that automatically generates CPPs by combining vision-based GUI understanding with privacy policy analysis. SeePrivacy detects privacy-relevant contexts with 0.88 precision and 0.90 recall, and extracts the corresponding policy segments with 0.98 precision and 0.96 recall. In a human evaluation, 77% of extracted segments were perceived as well-aligned with the detected contexts. These results suggest that SeePrivacy can significantly strengthen users’ interaction with, and understanding of, privacy policies, while making notices more accessible and inclusive for a broader audience. We then implements the framework as the first deployable Android SDK for contextual privacy policy generation. By elevating privacy disclosure from static documents to UI-level notices that can inform design, compliance checks, and developer tooling, we believe our work connects usable privacy engineering to software engineering practice. The proposed talk is based on a conference paper published in 33rd USENIX Security Symposium (USENIX Security 2024) Pa24 and a tool demonstration paper accepted by the 40th IEEE/ACM International Conference on Automated Software Engineering (ASE 2025) Go25.
Tao et al. (Thu,) studied this question.