Meta-Watermarking for Detection Recovery under Attack Shift: Empirical Analysis of FOMAML-Based Fast Adaptation

Diffusion-model watermark detectors are usually evaluated under fixed attack distributions, although deployed systems face continual attack shift. When out-of-distribution attacks reduce detection quality, full retraining is too slowfor latency-sensitive verification services. This paper formulateswatermark detection recovery as a task-level adaptation problem and proposes a meta-watermarking framework that learns a shared detector initialization via first-order Model-Agnostic Meta-Learning (FOMAML) on 10-dimensional ROBIN ring-frequency features. Recovery is measured by the minimum adaptation step k∗ needed to satisfy a fixed TPR/FPR operating point. Across six data scales (16–512 images), seven attack types, and four methods, the meta-initialized 11-parameter logistic detector achieves zero-shot recovery (k∗ = 0) on unseen rotation attacks at N = 128, with TPR=0.875 and FPR=0.063, whereas full retraining requires k∗ = 2. Threshold-only recalibration fails at this scale on rotation, indicating that learned initialization remains beneficial at moderate data budgets. After feature extraction, the detector classifies in 0.079 ms per image, about 43,000× smaller than the shared DDIM inversion stage that dominates end-to-end latency. Exploratory Spearman trends over the five single-seed scales (N = 16–256) suggested faster recovery with scale, but including the N = 512 stress test (n = 6) removes this correlation, and the multi-seed recovery rates are non-monotonic, so the trend does not generalize across all tested scales. The results are therefore consistent with meta-initialization as a practical recovery mechanism for the tested rotation attack shift, while recovery under high-noise conditions remains limited.