Reproduce “Watermark Overwriting Attack on StegaStamp algorithm” 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

I.F.Serzhenko, L.A.Khaertdinova, M.A.Pautov, A.V.Antsiferova

A "Watermark Overwriting Attack" was devised to neutralize the StegaStamp algorithm's embedded watermarks by extracting the original hidden message, inverting it, and re-embedding the inverted message. This technique achieved complete watermark removal (A=0) with minimal image quality degradation (Q=0.130), securing 5th place in the NeurIPS 2024 "Erasing the Invisible" competition.

Key insights

  • Sophisticated watermarking algorithms, despite being designed for robustness, can be vulnerable to attacks that leverage their own embedding and decoding mechanisms.
  • Overwriting an existing watermark by re-embedding its inverted binary message provides an effective strategy for complete removal.
  • It is possible to achieve complete watermark removal with minimal degradation of perceived image quality, posing a challenge for reliable content authentication.

Summary by alphaXiv.

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