GHCW: A novel Guarded High-fidelity Compression-based Watermarking scheme for AI model protection and self-recovery

Open Webpage

Tong Liu, Xiaochen Yuan, Wei Ke, Chan-Tong Lam, Sio Kei Im, Xiuli Bi

Published: 2025, Last Modified: 04 Nov 2025Appl. Soft Comput. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We propose a white-box fragile watermarking for tampered AI model self-recovery.•We employ SPIHT to compress model, enabling the generation of recovery watermark.•We protect the embedded recovery bits with linear encryption.