Automated detection of security sensitive UI elements

Automated detection of security sensitive UI elements for Clickshield

Designed a three-stage pipeline to automatically identify and tag security-sensitive UI widgets for Clickshield. Proposed using UI-CTX for static code analysis to construct UI Handler Graphs, applying a Graph Neural Network for binary classification at the graph level, and defining automatic code-level annotations (e.g., custom XML attributes) to drive runtime overlay protection.

Here is the full proposal.