Published on:
8 February 2024
Primary Category:
Multimedia
Paper Authors:
Ryoto Kanegae,
Masaki Kawamura
Proposes associative watermarking extending zero-watermarking
Uses hetero-associative memory for feature-to-watermark mapping
Uses auto-associative memory for watermark error correction
Demonstrates improved accuracy over zero-watermarking
Analyzes performance theoretically and empirically
Associative Watermarking for Image Authentication
This paper proposes an associative watermarking method that extends zero-watermarking for images. It applies associative memory models, specifically a hetero-associative model for mapping image features to watermarks, and an auto-associative model for watermark error correction. Through theory and simulations, the proposed model is shown to outperform zero-watermarking in terms of watermark retrieval accuracy. Its performance is analyzed using state equations and bit error rate.
Recovering lost watermarks using image denoising
Topic-based watermarking to identify AI text
Detecting and evaluating watermarks in AI text generation
Probabilistic verification of neural network ownership
Deep visual and audio watermarking for AI video editing forensics
Uncovering the Hidden Signals: A Popular Science Guide to Language Model Watermarking
No comments yet, be the first to start the conversation...
Sign up to comment on this paper