Reproduce “Detecting Generated-Depth Injection Attacks on RGB-D Place Authentication” 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
Seigo Kawachi, Satoshi Matsuura
The authors evaluate a place-authentication PAD defense against depth generated from public RGB and injected into the software path. The method combines reprojection error, which detects intrinsics mismatch within a view, with registration residual, which measures cross-view geometric disagreement. VGGT reaches 0.999 AUC with estimated intrinsics, but 0.817 after attacker-matched rounding and classifier retraining, in out-of-fold, not held-out, validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.