赞
踩
前言:本文介绍近5年来生成式跨模态隐写领域的相关工作。
相关阅读:生成式文本隐写发展综述
不同于文本隐写,跨模态隐写需要考虑不同模态间的相关性,常见的跨模态场景有:Image-to-Text(如图像描述), Text-to-Speech(如语音助手), Text-to-Image(如按文作画)等。下面对基于深度学习的生成式跨模态隐写相关工作进行介绍。
[1]- 基于图像描述的文本信息隐藏 (北京邮电大学学报,2018) BUPT, Xue et al.
bpw
;语义相关性:BLEU-N
[2]- Rits: real-time interactive text steganography based on automatic dialogue model (ICCCS, 2018) Tshinghua University, Yang et al.
这篇文章虽然不是跨模态的文章,但它指出生成的隐写文本应具备认知不可感知性,即:其语义应与上下文的语义相关,这一观点在跨模态文本隐写领域同样适用。
time
[3]- Steganographic visual story with mutual-perceived joint attention (EURASIP, 2021) Shanghai University, Guo et al.
Perplexity
;认知不可感知性:BLEU
&METEOR
[4]- ICStega: Image Captioning-based Semantically Controllable Linguistic Steganography (ICASSP, 2023) USTC, Wang et al.
bpw
;视觉不可感知性:Perplexity
;安全性:抗隐写分析能力 TS-FCN;认知不可感知性:BLEU
&METEOR
;多样性:LSA
&Self-CIDEr
[5]- Cross-Modal Text Steganography Against Synonym Substitution-Based Text Attack (SPL, 2023) Fudan University, Peng et al.
KL散度
;抗隐写分析能力:LS-CNN
&R-BIC
&SeSy
&BERT-FT
[6]- Cover Reproducible Steganography via Deep Generative Models (TDSC, 2022) USTC, Chen et al.
[7]- Distribution-Preserving Steganography Based on Text-to-Speech Generative Models (TDSC, 2022) USTC, Chen et al.
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。