Generative Steganography Scheme Based on StarGAN

doi:10.3969/j.issn.1671-1122.2020.12.009

Abstract

Abstract:

Aiming at the problems that the generated image and the real image are different in the generative steganography, and the image translation steganography needs to train a large number of models, a generative image steganography scheme based on StarGAN is proposed. Only one model can complete multi-style image translation task. The sender encodes the secret information, maps it to the style tag of the image, generates an image of the corresponding style, and sends it to the receiver. The receiver uses the extraction model which passed by the secret channel to extract the style tag of the image, and compares the encoding method to decode the secret informatione. The experimental results show that while reducing the number of training models, the scheme has significantly improved image quality and information extraction accuracy.

Key words: information hiding, deep learning, generative steganography, image translation

CLC Number:

TP309

BI Xinliang, YANG Haibin, YANG Xiaoyuan, HUANG Siyuan. Generative Steganography Scheme Based on StarGAN[J]. Netinfo Security, 2020, 20(12): 64-71.

Figures/Tables 11

References 21

[1]	LUO Weiqi, HUANG Fangjun, HUANG Jiwu. Edge Adaptive Image Steganography Based on LSB Matching Revisited[J]. IEEE Transactions on Information Forensics & Security, 2010,5(2):201-214.
[2]	FILLER T, JUDAS J, FRIDRICH J. Minimizing Additive Distortion in Steganography Using Syndrome-Trellis Codes[J]. IEEE Transactions on Information Forensics & Security, 2011,6(3):920-935.
[3]	GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative Adversarial Networks[EB/OL]. https://arxiv.org/abs/1406.2661, 2014-06-10.
[4]	ZHANG Zhu, LIU Jia, KE Yan, et al. Generative Steganography by Sampling[J]. IEEE Access, 2019,19(7):118586-118597.
[5]	HUSSAIN I, ZENG Jishen, TAN Sunquan. A Survey on Deep Convolutional Neural Networks for Image Steganographyand Steganalysis[J]. KSII Trans on Internet and Information Systems, 2020,14(3):1228-1248.
[6]	VOLKHONSKIY D, BORISENKO B, BURNAEV E. Generative Dversarial Networks for Image Steganography[EB/OL]. https://openreview.net/pdf?id=H1hoFU9xe, 2020-08-09.
[7]	ZHU Junyan, PARK T, ISOLA P, et al. Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks[C]// IEEE. Proceedings of the IEEE International Conference on Computer Vision, October 22-29, 2017, Venice, Italy. NJ: IEEE, 2017: 2223-2232.
[8]	HAYES J, DANEZIS G. Generating Steganographic Images via Adversarial Training[EB/OL]. https://arxiv.org/abs/1703.00371, 2014-7-13.
[9]	HU Donghui, WANG Liang, JIANG Wenjie, et al. A Novel Image Steganography Method via Deep Convolutional Generative Adversarial Networks[J]. IEEE Access, 2018,18(6):38303-38314.
[10]	ZHANG Zhuo, FU Guangyuan, NI Rongrong, et al. A Generative Method for Steganography by Cover Synjournal with Auxiliary Semantics[J]. Tsinghua Science and Technology, 2020,25(4):516-527. doi: 10.1109/TST.5971803 URL
[11]	TANG Weixuan, TAN Shunquan, LI Bin, et al. Automatic Steganographic Distortion Learning Using a Generative Adversarial Network[J]. IEEE Signal Processing Letters, 2017,24(10):1547-1551. doi: 10.1109/LSP.2017.2745572 URL
[12]	YANG Jianhua, LIU Kai, KANG Xiangui, et al. Spatial Image Steganography Based on Generative Adversarial Network[EB/OL]. https://arxiv.org/abs/1804.07939, 2018-04-21.
[13]	RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional Networks for Biomedical Image Segmentation[C]// TUM & FAU. International Conference on Medical Image Computing and Computer-assisted Intervention, October 5-9, 2015, Munich, Germany. Cham: Springer, 2015: 234-241.
[14]	HOLUB V, FRIDRICH J, DENEMARK T. Universal Distortion Function for Steganography in an Arbitrary Domain[J]. EURASIP Journal on Information Security, 2014,14(1):1.
[15]	CHOI Y, CHOI M, KIM M, et al. Stargan: Unified Generative Adversarial Networks for Multi-Domain Image-to-Imagetranslation[C]// IEEE. Proc of the IEEE Conference on Computer Vision and Pattern Recognition, June 18-22, 2018, Salt Lake City, Utah, USA. NJ: IEEE, 2018: 8789-8797.
[16]	PERARNAU G, VAN D J, RADUCANU B, et al. Invertible Conditional Gans for Image Editing[EB/OL]. https://arxiv.org/abs/1611.06355, 2016-11-19.
[17]	LIU Jia, KE Yan, ZHANG Zhuo, et al. Recent Advances of Image Steganography with Generative Adversarial Networks[J]. IEEE Access, 2020,20(8):60575-60597.
[18]	HICSONMEZ S, SAMET N, AKBAS E, et al. GANILLA: Generative Adversarial Networks for Image to Illustration Translation[EB/OL]. https://arxiv.org/abs/2002.05638, 2020-02-14.
[19]	ISOLA P, ZHU Junyan, ZHOU Tinghui, et al. Image-to-Image Translation with Conditional Adversarial Networks[C]// IEEE. Proc of The IEEE Conference on Computer Visionan Dpattern Recognition, July 21-26, 2017, Honolulu, Hawaii, USA. NJ: IEEE, 2017: 1125-1134.
[20]	HEUSEL M, RAMSAUER H, UNTERTHINER T, et al. Gans Trainedbya Two Time-Scale Update Rule Converge to ALocal Nash Equilibrium[EB/OL]. https://arxiv.org/abs/1706.08500v6, 2018-01-12.
[21]	KE Yan, LIU Jia, ZHANG Minqing, et al. Steganography Security: Princip leand Practice[J]. IEEE Access, 2018,18(6):73009-73022.

	大嘴唇	双下巴	山羊胡子	年轻	···	刘海
风格标签码	1	-1	-1	1	···	-1
秘密消息	1	0	0	1	···	0
风格标签码	1	-1	-1	1	···	-1
秘密消息	1	0	0	1	···	0
风格标签码	1	-1	-1	1	···	1
秘密消息	1	0	0	1	···	1
风格标签码	-1	-1	-1	1	···	-1
秘密消息	0	0	0	1	···	0

网络名称	转换风格类别	需要模型数
IcGAN	2	2
CycleGAN	2	2
本文方案	2	1
IcGAN	3	6
CycleGAN	3	6
本文方案	3	1

网络名称	FID值
IcGAN	155.5442
CycleGAN	45.3995
本文方案	34.6783

方法	提取准确率
文献[9]	89.3%-96.0%
本文方案（64×64）	88.6%
本文方案（128×128）	95.0%