Reproduce “Sequential Semantic-Non-Semantic Watermarking: Robustness Against UnMarker Attacks” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Md Ifraham Iqbal

The authors test whether embedding two independently decoded source signatures in one image can make watermark removal harder. They combine HiDDeN, treated as high-frequency, with StegaStamp, treated as lower-frequency, and evaluate UnMarker attacks on 50 images. In a representative sequence, one attack leaves StegaStamp bit accuracy at 0.988; two attacks lower both scores, but do not establish a general defense.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.