基于非线性残差的JPEG图像隐写分析

doi:10.11959/j.issn.1000-436x.2023010

通信学报 ›› 2023, Vol. 44 ›› Issue (1): 142-152.doi: 10.11959/j.issn.1000-436x.2023010

基于非线性残差的JPEG图像隐写分析

夏超¹, 刘亚奇¹, 关晴骁², 金鑫¹, 张艳硕¹, 许盛伟¹

¹ 北京电子科技学院，北京 100070
² 集美大学计算机工程学院，福建厦门 361021

修回日期:2022-10-31 出版日期:2023-01-25 发布日期:2023-01-01
作者简介:夏超（1987- ），男，浙江上虞人，博士，北京电子科技学院助理研究员，主要研究方向为信息隐藏
刘亚奇（1991- ），男，山西晋城人，博士，北京电子科技学院工程师，主要研究方向为多媒体安全、图像处理
关晴骁（1984- ），男，湖南湘潭人，博士，集美大学副教授，主要研究方向为信息隐藏、多媒体安全
金鑫（1983- ），男，安徽安庆人，博士，北京电子科技学院副教授，主要研究方向为人工智能与安全
张艳硕（1979- ），男，陕西宝鸡人，博士，北京电子科技学院副教授，主要研究方向为密码算法设计、密码协议分析、区块链技术
许盛伟（1976- ），男，江西吉安人，博士，北京电子科技学院研究员，主要研究方向为网络安全
基金资助:
国家自然科学基金资助项目(62002003);国家自然科学基金资助项目(U1936114);国家自然科学基金资助项目(62102010);北京高校“高精尖”学科建设基金资助项目(20210085Z0401);北京高校“高精尖”学科建设基金资助项目(20210037Z0401)

Steganalysis of JPEG images using non-linear residuals

Chao XIA¹, Yaqi LIU¹, Qingxiao GUAN², Xin JIN¹, Yanshuo ZHANG¹, Shengwei XU¹

¹ Beijing Electronic Science and Technology Institute, Beijing 100070, China
² Computer Engineering College, Jimei University, Xiamen 361021, China

Revised:2022-10-31 Online:2023-01-25 Published:2023-01-01
Supported by:
The National Natural Science Foundation of China(62002003);The National Natural Science Foundation of China(U1936114);The National Natural Science Foundation of China(62102010);Advanced Discipline Construction Project of Beijing Universities(20210085Z0401);Advanced Discipline Construction Project of Beijing Universities(20210037Z0401)

摘要/Abstract

摘要：

当前JPEG图像隐写分析方法往往只使用线性滤波器生成残差图像，导致残差图像类型过于单一，影响隐写检测性能。鉴于此，提出了基于非线性残差的JPEG图像隐写分析方法。首先，从现有JPEG图像隐写分析方法已有的线性残差图像中选择合适的图像组，对其采用逐像素取小和取大运算，在没有明显增加计算复杂度的情况下，得到非线性残差图像。其次，根据JPEG相位，将非线性残差图像划分为子图像，提取子图像直方图特征。再次，充分考虑非线性运算的影响，设计合适的特征对称合并方法。最后，将合并后的直方图特征拼接形成最终的隐写分析特征。实验结果表明，线性和非线性残差图像相结合能够有效地增加残差图像和隐写分析特征的多样性，从而提升JPEG图像隐写检测性能。

关键词: 信息隐藏, 隐写分析, 隐写, JPEG图像

Abstract:

Most current JPEG steganalytic methods can only extract features from a series of linear residuals.Non-linear filters are not considered in these JPEG steganalytic methods, resulting in single types of residuals.Hence, a JPEG steganalytic method using non-linear residuals was proposed.Firstly, non-linear residuals were generated without a high computational cost by using element-wise minimum and maximum operations across a couple of linear residuals which had been obtained in the current JPEG steganalytic method.Secondly, according to the JPEG phase, the non-linear residual was divided into sub-residuals in which the histogram features were extracted.Thirdly, considering the minimum and maximum operators, the symmetrization method was accordingly designed.Finally, all the symmetrized histogram features were concatenated to form the final feature set.Experimental results indicate that the performance for JPEG steganalysis can be improved effectively by using both the linear and the non-linear residuals.

Key words: information hiding, steganalysis, steganography, JPEG image

中图分类号:

TP309

夏超, 刘亚奇, 关晴骁, 金鑫, 张艳硕, 许盛伟. 基于非线性残差的JPEG图像隐写分析[J]. 通信学报, 2023, 44(1): 142-152.

Chao XIA, Yaqi LIU, Qingxiao GUAN, Xin JIN, Yanshuo ZHANG, Shengwei XU. Steganalysis of JPEG images using non-linear residuals[J]. Journal on Communications, 2023, 44(1): 142-152.

图/表 7

图1

表1

DCTR的非线性滤波策略中的$lt@mathml$gt@6E3FFB21-324E-40D4-8B91-563910AAC57C.ml$lt@/mathml$gt@"

c					d
c	0	1	2	3	4	5	6	7
0	0.023	0.043	0.034	0.056	0.047	0.057	0.030	0.037
1	0.036	0.051	$0 . 000$	0.057	0.033	0.055	$0 . 004$	0.040
2	0.030	$0 . 002$	0.039	0.032	0.018	0.033	0.027	$0 . 009$
3	0.040	0.038	0.023	0.043	0.023	0.042	0.015	0.041
4	0.050	0.033	0.020	0.025	0.029	0.025	0.021	0.033
5	0.040	0.044	0.019	0.042	0.024	0.041	0.021	0.040
6	0.030	$0 . 008$	0.029	0.028	0.018	0.035	0.036	$0 . 003$
7	0.027	0.039	$0 . 009$	0.046	0.028	0.050	$- 0 . 001$	0.040

表1

表2

GFR的非线性滤波策略中的 $lt@mathml$gt@B86B5C4E-9035-4629-9A59-F5514494972F.ml$lt@/mathml$gt@"

c					d
c	0	1	2	3	4	5	6	7
0	0.023	0.019	0.010	0.061	0.055	0.060	0.012	0.019
1	0.022	0.030	$0 . 000$	0.086	0.061	0.068	$- 0 . 003$	0.026
2	0.017	$0 . 000$	0.027	0.060	0.060	0.053	0.029	$0 . 000$
3	0.057	0.056	0.062	0.065	0.068	0.071	0.049	0.053
4	0.060	0.053	0.062	0.069	0.054	0.063	0.058	0.047
5	0.051	0.050	0.049	0.070	0.060	0.070	0.047	0.050
6	0.018	$0 . 000$	0.028	0.050	0.059	0.051	0.022	0.001
7	0.019	0.023	$- 0 . 003$	0.069	0.052	0.069	$- 0 . 003$	0.022

表2

表3

原始DCTR、DCTRmnmx、DCTRspam+mnmx的检测错误率"

隐写分析特征	特征维数	滤波方法	${\bar{P}}_{E}$
原始DCTR	8 000	线性滤波	0.153 9
DCTRmnmx	2 100	非线性滤波	0.159 8
DCTRspam+mnmx	10 100	线性＆非线性滤波	$0 . 1433$

表3

表4

原始PHARM、PHARMmnmx、PHARMspam+mnmx的检测错误率"

隐写分析特征	特征维数	滤波方法	${\bar{P}}_{E}$
原始PHARM	12 600	线性滤波	0.130 4
PHARMmnmx	12 600	非线性滤波	0.128 6
PHARMspam+mnmx	25 200	线性＆非线性滤波	$0 . 1182$

表4

表5

原始GFR、GFRmnmx、GFRspam+mnmx、GRF的检测错误率"

隐写分析特征	特征维数	滤波方法	${\bar{P}}_{E}$
原始GFR	17 000	线性滤波	0.103 1
GFRmnmx	5 400	非线性滤波	0.101 7
GFRspam+mnmx	22 400	线性＆非线性滤波	$0 . 0945$
GRF	92 412	线性滤波	0.095 2

表5

表6

多种隐写分析特征在不同条件下的检测错误率"

质量因子	隐写算法	嵌入率/bpnzAC	原始DCTR	DCTRspam+mnmx	原始PHARM	PHARMspam+mnmx	原始GFR	GFRspam+mnmx
		0.05	0.464 1	0.459 6	0.459 8	0.456 1	0.450 1	0.448 7
		0.1	0.411 8	0.406 2	0.388 0	0.385 5	0.383 4	0.376 6
	J-UNIWARD	0.2	0.296 0	0.281 2	0.241 6	0.238 4	0.238 9	0.228 4
		0.3	0.186 1	0.171 7	0.131 9	0.128 0	0.131 2	0.122 9
55		0.4	0.107 1	0.096 4	0.066 9	0.062 3	0.066 9	0.060 8
55		0.05	0.461 9	0.455 5	0.454 0	0.448 7	0.445 3	0.440 2
		0.1	0.404 7	0.396 0	0.371 7	0.365 8	0.374 0	0.363 2
	UERD	0.2	0.288 5	0.274 7	0.227 6	0.223 4	0.237 7	0.226 0
		0.3	0.185 0	0.171 1	0.123 1	0.119 5	0.138 0	0.129 6
		0.4	0.107 8	0.097 3	0.067 2	0.062 7	0.074 5	0.068 2
		0.05	0.478 1	0.474 4	0.474 3	0.472 0	0.464 1	0.462 1
		0.1	0.437 8	0.434 3	0.429 3	0.424 0	0.410 8	0.405 6
	J-UNIWARD	0.2	0.338 9	0.331 4	0.319 5	0.309 0	0.2871	0.278 2
		0.3	0.238 2	0.227 7	0.213 1	0.198 7	0.179 5	0.169 1
75		0.4	0.153 9	0.143 3	0.130 4	0.118 2	0.103 1	0.094 5
75		0.05	0.468 5	0.465 8	0.469 1	0.465 8	0.464 2	0.461 7
		0.1	0.425 5	0.420 7	0.424 3	0.419 0	0.415 9	0.409 4
	UERD	0.2	0.331 3	0.324 8	0.326 8	0.316 0	0.309 9	0.300 3
		0.3	0.246 1	0.237 2	0.237 3	0.223 6	0.217 0	0.209 6
		0.4	0.173 7	0.166 2	0.169 5	0.155 7	0.147 4	0.137 2
		0.05	0.496 0	0.494 6	0.495 1	0.495 0	0.492 5	0.491 9
		0.1	0.484 7	0.480 1	0.482 8	0.481 0	0.475 1	0.473 9
	J-UNIWARD	0.2	0.440 5	0.434 2	0.442 6	0.436 0	0.422 4	0.418 2
		0.3	0.379 8	0.369 5	0.383 9	0.372 2	0.350 8	0.343 5
95		0.4	0.310 0	0.299 1	0.312 8	0.292 1	0.269 4	0.261 7
95		0.05	0.494 4	0.492 2	0.495 2	0.493 2	0.491 4	0.491 3
		0.1	0.481 5	0.476 8	0.483 7	0.480 9	0.473 4	0.472 9
	UERD	0.2	0.437 5	0.430 7	0.443 6	0.442 5	0.427 9	0.424 5
		0.3	0.380 5	0.372 1	0.395 2	0.387 4	0.369 4	0.363 5
		0.4	0.313 2	0.305 1	0.334 6	0.322 4	0.299 2	0.289 7

表6

参考文献 25

[1]	FILLER T , JUDAS J , FRIDRICH J . Minimizing additive distortion in steganography using syndrome-trellis codes[J]. IEEE Transactions on Information Forensics and Security, 2011,6(3): 920-935.
[2]	HOLUB V , FRIDRICH J , DENEMARK T . Universal distortion function for steganography in an arbitrary domain[J]. EURASIP Journal on Information Security, 2014(1): 1-13.
[3]	GUO L J , NI J Q , SU W K ,et al. Using statistical image model for JPEG steganography:uniform embedding revisited[J]. IEEE Transactions on Information Forensics and Security, 2015,10(12): 2669-2680.
[4]	汤光明, 孙艺, 徐潇雨 ,等. 动态更新失真代价的自适应 JPEG 隐写算法[J]. 通信学报, 2017,38(9): 1-8.
	TANG G M , SUN Y , XU X Y ,et al. Adaptive JPEG steganography based on distortion cost updating[J]. Journal on Communications, 2017,38(9): 1-8.
[5]	SU W K , NI J Q , LI X H ,et al. A new distortion function design for JPEG steganography using the generalized uniform embedding strategy[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2018,28(12): 3545-3549.
[6]	HU X L , NI J Q , SHI Y Q . Efficient JPEG steganography using domain transformation of embedding entropy[J]. IEEE Signal Processing Letters, 2018,25(6): 773-777.
[7]	CHEN K J , ZHOU H , ZHOU W B ,et al. Defining cost functions for adaptive JPEG steganography at the microscale[J]. IEEE Transactions on Information Forensics and Security, 2019,14(4): 1052-1066.
[8]	WANG Y F , LI W X , ZHANG W M ,et al. BBC:enhanced block boundary continuity on defining non-additive distortion for JPEG steganography[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2021,31(5): 2082-2088.
[9]	WANG Z C , FENG G R , QIAN Z X ,et al. JPEG steganography with content similarity evaluation[J]. IEEE Transactions on Cybernetics, 2022:doi.org/10.1109/TCYB.2022.3155732.
[10]	PEVNY T , FRIDRICH J . Merging Markov and DCT features for multi-class JPEG steganalysis[C]// Proceeding of Security,Steganography,and Watermarking of Multimedia Contents IX. Bellingham:SPIE Press, 2007: 28-40.
[11]	KODOVSK J , FRIDRICH J . Steganalysis of JPEG images using rich models[C]// Proceeding of Media Watermarking,Security,and Forensics. Bellingham:SPIE Press, 2012: 81-93.
[12]	HOLUB V , FRIDRICH J . Low-complexity features for JPEG steganalysis using undecimated DCT[J]. IEEE Transactions on Information Forensics and Security, 2015,10(2): 219-228.
[13]	HOLUB V , FRIDRICH J . Phase-aware projection model for steganalysis of JPEG images[C]// Proceeding of Media Watermarking,Security,and Forensics. Bellingham:SPIE Press, 2015: 259-269.
[14]	SONG X F , LIU F L , YANG C F ,et al. Steganalysis of adaptive JPEG steganography using 2D Gabor filters[C]// Proceedings of the 3rd ACM Workshop on Information Hiding and Multimedia Security. New York:ACM Press, 2015: 15-23.
[15]	FRIDRICH J , KODOVSKY J . Rich models for steganalysis of digital images[J]. IEEE Transactions on Information Forensics and Security, 2012,7(3): 868-882.
[16]	LI B , LI Z P , ZHOU S J ,et al. New steganalytic features for spatial image steganography based on derivative filters and threshold LBP operator[J]. IEEE Transactions on Information Forensics and Security, 2018,13(5): 1242-1257.
[17]	XIA C , GUAN Q X , ZHAO X F ,et al. Improving GFR steganalysis features by using Gabor symmetry and weighted histograms[C]// Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security. New York:ACM Press, 2017: 55-66.
[18]	XIA C , GUAN Q X , ZHAO X F ,et al. Improved JPEG phase-aware steganalysis features using multiple filter sizes and difference images[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2020,30(11): 4100-4113.
[19]	BAS P , FILLER T , PEVNY T , . “Break our steganographic system”:the ins and outs of organizing BOSS[C]// Proceedings of the 13th International Workshop on Information Hiding. Berlin:Springer, 2011: 59-70.
[20]	KODOVSKY J , FRIDRICH J , HOLUB V . Ensemble classifiers for steganalysis of digital media[J]. IEEE Transactiions on Informatioon Forensics andd Security, 2012,7(2): 432-444.
[21]	SONG X F , LIU F L , ZHANG Z G ,et al. 2D Gabor filters-based steganalysis of content-adaptivve JPEG steganogrraphy[J]. Multimeddia Tools and Appplications, 2017,776(24): 26391-264119.
[22]	YE J , NI J Q , YI Y . Deep leearning hierarchicaal representations ffor image stegannalysis[J]. IEEE Transactions on Innformation Forensiics and Security, 2017,12(11): 2545-2557.
[23]	史晓裕, 李斌, 谭舜泉 . 深度学习空域隐写分析的预处理层[J]. 应用科学学报, 2018,36(2): 309-3320.
	SHI X Y , LI B , TAN S Q . Preprocessing layer inn spatial steganalyssis based on deeep learning[J]. Journal of Applied Scciences, 2018,36(22): 309-320.
[24]	王晓丹, 李京泰, 宋亚飞 . DDAC:面向卷积神经网络图像隐写分析模型的特征提取方法[J]. 通信学报, 2022,43((5): 68-81.
	WANG X D , LI J T , SONG Y F . DDAC:a featurre extraction methood for model off image steganalyssis based on convoolutional neural neetwork[J]. Jourrnal on Communicaations, 2022,43(5): 68-81.
[25]	YANG J H , KANG X G , WONG E K ,et al. JPEG steganalysis wiith combined dense connected CNNs and SCA-GFR[J]. Multimedia Toools and Applicatiions, 2019,78(7): 88481-8495.

基于非线性残差的JPEG图像隐写分析

Steganalysis of JPEG images using non-linear residuals

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 25

相关文章 15

Metrics

推荐阅读 0

[1]	刘玉玲, 王翠林, 付章杰. 语义空间下基于情感表达的生成式文本隐写方法[J]. 通信学报, 2023, 44(4): 176-186.
[2]	黄冬艳, 李琨. 多地址的时间型区块链隐蔽通信方法研究[J]. 通信学报, 2023, 44(2): 148-159.
[3]	李晖, 金家立, 金纾羽, 马卫娇. 基于自动选择编码及动态选词策略的文本隐写方法[J]. 通信学报, 2022, 43(9): 240-253.
[4]	吴友情, 马文静, 殷赵霞, 彭银银, 张新鹏. 基于预测误差位平面压缩的密文图像可逆信息隐藏[J]. 通信学报, 2022, 43(8): 219-230.
[5]	王晓丹, 李京泰, 宋亚飞. DDAC：面向卷积神经网络图像隐写分析模型的特征提取方法[J]. 通信学报, 2022, 43(5): 68-81.
[6]	佘维, 荣欣鹏, 刘炜, 田钊. 基于马尔可夫链的生成式区块链隐蔽通信模型[J]. 通信学报, 2022, 43(10): 121-132.
[7]	杨晓元, 毕新亮, 刘佳, 黄思远. 结合图像加密与深度学习的高容量图像隐写算法[J]. 通信学报, 2021, 42(9): 96-105.
[8]	周志立, 王美民, 杨高波, 朱剑宇, 孙星明. 基于轮廓自动生成的构造式图像隐写方法[J]. 通信学报, 2021, 42(9): 144-154.
[9]	张祯, 倪嘉铭, 姚晔, 龚礼春, 王玉娟, 吴国华. 基于同义词扩展和标签传递机制的文本无载体信息隐藏方法[J]. 通信学报, 2021, 42(9): 173-183.
[10]	竺乐庆,郭钰,莫凌强,张大兴. DGANS：基于双重生成式对抗网络的稳健图像隐写模型[J]. 通信学报, 2020, 41(1): 125-133.
[11]	吴国华,龚礼春,袁理锋,姚晔. 中文文本信息隐藏研究进展[J]. 通信学报, 2019, 40(9): 145-156.
[12]	马利民,王佳慧. 基于改进FEMD算法的可逆秘密图像共享方案[J]. 通信学报, 2019, 40(7): 48-56.
[13]	任帅,王震,苏东旭,张弢,慕德俊. 基于三维模型贴图与结构数据的信息隐藏算法[J]. 通信学报, 2019, 40(5): 211-222.
[14]	姚远志,王锋,严文博,俞能海. 基于二维码和可逆可视水印的图像隐私保护方案[J]. 通信学报, 2019, 40(11): 65-75.
[15]	汤光明,孙艺,徐潇雨,王宇. 动态更新失真代价的自适应JPEG隐写算法[J]. 通信学报, 2017, 38(9): 1-8.