摘 要：图像隐写是一种重要的隐蔽通信技术。现有彩色图像隐写算法未能在容量和安全性之间达到良好平衡。针对上述问题，通过引入通道注意力机制，使嵌入修改集中在纹理复杂区域以改善生成式隐写算法安全性较低的问题。利用彩色图像不同颜色通道的关联性设计复合损失函数，提高基于生成对抗网络的彩色图像隐写性能。采用公开彩色图像数据集 COCO 进行实验。实验结果表明，所提算法在高嵌入率情况下安全性优于现有彩色图像隐写算法。
关键词：彩色图像隐写；注意力机制；生成对抗网络；安全性
DOI:10.19850/j.cnki.2096-4706.2022.07.018
中图分类号：TP309 文献标识码：A 文章编号：2096-4706（2022）07-0070-07
A Color Image Steganography Algorithm Based on Attention Mechanism and Generative Adversarial Network
CHEN Menghua, LIU Jiayong, HE Peisong
(School of Cyber Science and Engineering, Sichuan University, Chengdu 610065, China)
Abstract: Image steganography is an important technique for covert communication. The existing color image steganography algorithms fail to achieve a good balance between capacity and security. To address this problem, this paper improves the problem of low security of the generative steganography algorithm by introducing a channel attention mechanism to focus the embedding modifications on the texture complex region. It designs composite loss function by using the correlation of different color channels of color images to improve the performance of color image steganography based on generative adversarial networks. It uses the open color image dataset COCO to experimentize. The experimental results show that the security of proposed algorithm outperforms existing color image steganography algorithm at high embedding rate situation.
Keywords: color image steganography; attention mechanism; generative adversarial network; security
参考文献：
[1] 谭艳萍，罗永，张俊 . 基于深度学习的数字图像隐写和隐写分析术研究概述 [J]. 现代信息科技，2021，5（13）：68-72.
[2] 张祎，罗向阳，王金伟，等 . 数字图像鲁棒隐写综述 [J].中国图象图形学报，2022，27（1）：3-26.
[3] ZHANG R，DONG S Q，LIU J Y. Invisible steganography via generative adversarial networks [J].Multimed Tools Appl，2019，78（7）：8559-8575.
[4] JING J，DENG X，XU M，et al. HiNet：Deep Image Hiding by Invertible Network [C]//Proceedings of the IEEE/CVF International Conference on Computer Vision （ICCV）.Montreal：IEEE，2021：4733-4742.
[5] SU H，YANG S，ZHANG S，et al. A Color Image Steganography Based on Frequency Sub-band Selection [J/OL].[2022-02-04].https://arxiv.org/abs/2112.14437.
[6] HU J，SHEN L，ALBANIE S. Squeeze-and-Excitation Networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2020，42（8）：2011-2023.
[7] LIAO X，YU Y，LI B，et al. A New Payload Partition Strategy in Color Image Steganography [J].IEEE Trans. Circuits Syst. Video Technol.，2020，30（3）：685-696.
[8] TANG W X，LI B，LUO W Q，et al. Clustering steganographic modification directions for color components [J].IEEE Signal Process Lett，2016，23（2）：197-201.
[9] QIN X，LI B，TAN S，et al. A Novel Steganography for Spatial Color Images Based on Pixel Vector Cost [J].IEEE Access， 2019，7：8834-8846.
[10] WANG Y，ZHANG W，LI W，et al. Non-Additive Cost Functions for Color Image Steganography Based on Inter-Channel Correlations and Differences [J].IEEE Trans. Inf. Forensics Secur.，2020，15：2081-2095.
[11] LI B，WANG M，LI X，et al. A Strategy of Clustering Modification Directions in Spatial [J].IEEE Trans. Inf. Forensics Secur.，2015，10（9）：1905-1917.
[12] ZHANG K A，CUESTA-INFANTE A，XU L，et al. Image SteganographySteganoGAN：High Capacity Image Steganography with GANs [J/OL].[2022-02-04].https: //arxiv.org/abs/1901.03892v2.
[13] LI Y，LIU J，LIU X，et al. HCISNet：Higher-capacity invisible image steganographic network [J].IET Image Process.，2021，15（13）：3332-3346.
[14] BALUJA S. Hiding Images in Plain Sight：Deep Steganography [C]//Advances in Neural Information Processing Systems 30：Annual Conference on Neural Information Processing Systems 2017.Long Beach：ACM，2017：2069-2079.
[15] ZHANG R，DONG S，LIU J. Invisible steganography via generative adversarial networks [J].Multim. Tools Appl.，2019，78（7）： 8559-8575.
[16] YANG J，RUAN D，HUANG J，et al. An Embedding Cost Learning Framework Using GAN [J].IEEE Trans. Inf. Forensics Secur.，2020，15：839-851.
[17] FRIDRICH J J，KODOVSKY J. Rich Models for Steganalysis of Digital Images [J].IEEE Trans. Inf. Forensics Secur.，2012，7（3）：868-882.
[18] XU G，WU H Z，SHI Y Q. Structural Design of Convolutional Neural Networks for Steganalysis [J].IEEE Signal Process. Lett.，2016，23（5）：708-712.
[19] LIN T Y，MAIRE M，BELONGIE S，et al. Microsoft COCO：Common Objects in Context [C]//Computer Vision–ECCV 2014. Cham：Springer International Publishing，2014：740-755.
[20] THE IMAGEMAGICK DEVELOPMENT TEAM. ImageMagick [EB/OL].[2022-02-04].https://imagemagick.org/.
作者简介：陈孟华（1997—），女，汉族，陕西西安人，硕士研究生在读，研究方向：信息内容安全、图像隐写；刘嘉勇（1962—），男，汉族，四川成都人，教授，博士，研究方向：信息安全理论与应用、网络通信与网络安全；通讯作者：何沛松（1991—），男，汉族，四川成都人，副教授，博士，研究方向：多媒体安全、人工智能。