Interactive web lab visualizing AI watermark robustness
The project provides a browser‑based teaching lab that injects synthetic watermarks into images and text, then lets users apply filters and edits to see how the signal degrades. It uses React, Vite, and TypeScript to run all signal‑processing locally, so no data leaves the user’s machine. Designed for AI‑safety researchers, educators, and developers interested in provenance, it demonstrates why cryptographic content credentials are needed beyond simple watermarks. Compared to generic demos, it offers real signal‑processing utilities and clear visualizations, making the concepts tangible.
View on GitHub →mizcausevic-dev/watermark-robustness-lab