Text Zero-watermarking Algorithm Based on Neural Machine Translation
DING Changhao
YU Qi
WANG Fan
Abstract：As an important way of copyright protection and authentication,text zero-watermarking can effectively detect whether text has been tampered with or embezzled.But now text zero-watermarking still faces the problems of low robustness and poor resistance to large-scale content modification attacks.To address the problems above,this paper proposes a method of generat-ing zero watermark based on neural machine translation(NMT)model.This method inputs the text that needs to be protected into the pre-trained NMT model,obtains the translation and information entropy value in the generation process,and sends them as zero wa-termark to a trusted third party for saving.In the copyright verification process,the method calculates the BLEU value and the infor-mation entropy difference between the two translations for similarity verification.This method can effectively resist attacks such as synonym replacement attack,content replacement,and text deletion,and improve the robustness of text zero-watermarking when attacked.
Keywords：text zero-watermarkingneural machine translationinformation entropy
Publication Date：2025-10-20
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：6( 2697-2702 )