计算机与数字工程2025，Vol.53Issue(10)：2697-2702,6.DOI:10.3969/j.issn.1672-9722.2025.10.005
基于神经机器翻译模型的文本零水印算法
Text Zero-watermarking Algorithm Based on Neural Machine Translation
DING Changhao 1YU Qi 1WANG Fan1
作者信息
- 1. Engineering Research Center of Digital Forensics,Ministry of Education,Nanjing University of Information Science&Technology,Nanjing 210044
- 折叠
摘要
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.关键词
文本零水印/神经机器翻译/信息熵Key words
text zero-watermarking/neural machine translation/information entropy分类
信息技术与安全科学引用本文复制引用
DING Changhao,YU Qi,WANG Fan..基于神经机器翻译模型的文本零水印算法[J].计算机与数字工程,2025,53(10):2697-2702,6.
