基于颜色梯度特性的彩色图像隐写分析

doi:1000-436X(2011)01-0027-10

摘要/Abstract

摘要：

基于彩色图像任意2个颜色通道间梯度方向随机序列及颜色梯度和随机序列的振荡特性在隐写前后均有明显变化的特征，提出了一种基于颜色梯度特性的彩色图像盲隐写分析算法。算法应用 HHT(Hilbert-Huang transform)提取4个序列在隐写前后的内在振荡特征差异，构建基于Hilbert谱的特征向量，采用SVM分类器进行分类判别。实验表明，该算法误检率低，检测性能优于现有彩色图像隐写分析算法，可以实现彩色图像隐写的可靠检测。

关键词: 彩色图像隐写分析, Mumford-Shah模型, 颜色梯度和序列, 颜色梯度方向序列, HHT

Abstract:

Based on the difference between the vibrations of color gradient orientation sequences and color gradient sum sequences of stego images and those of images without iding message,a novel reliable steganalysis of color images was presented which major in the analysis of these two categories of color gradient sequences,which could distill the differences of the vibrations of them.The Hilbert-Huang transform based characteristic vectors were constructed and the SVM classifier was used to classification.The experimenta results have shown that the proposed method realizes the reliable steganalysis of color images with higher correct rate,shorter vectors and lower false positive rate comparing with trad itional color image steganalysis algorithms.

Key words: color image steganalysis;Mumford-Shah model, color gradient sum sequence, color gradient orientation sequence, Hilbert-Huang transform

綦科,张大方,谢冬青. 基于颜色梯度特性的彩色图像隐写分析[J]. 通信学报, 2011, 32(1): 27-36.

Ke QI,Da-fang ZHANG,Dong-qing XIE. Reliable steganalysis of color images based on color gradient sequence[J]. Journal on Communications, 2011, 32(1): 27-36.

图/表 7

图1

图2

图3

图4

表1

表2

表3

参考文献 22

[1]	罗向阳, 陆佩忠, 刘粉林 . 一类可以抵御SPA分析的动态补偿LSB信息隐藏方法[J]. 计算机学报, 2007,30(3): 463-473. LUO X Y , LU P Z , LIU F L . A dynamic compensation LSB steganography method defeating SPA[J]. Chinese Journal of Computer, 2007,30(3): 463-473.
[2]	FRIDRICH J , GOLJAN M . Digital image steganography using stochastic modulation[A]. Proceedings of SPIE[C]. Santa Clara,CA, 2003. 191-203.
[3]	WANG K S , KIYOSHI T , QI X J . Multiple messages embedding using DCT-based Mod4 steganographic method[A]. MRCS 2006[C]. LNCS 4105, 2006. 57-65.
[4]	SALLEE P . Model-based methods for steganography and steganalysis[J]. International Journal of Image and Graphics, 2005,5(1): 167-189.
[5]	GAO H Y,XU Y , LI X , LIU G Q . A steganographic algorithm for JPEG 2000 image[A]. IEEE International Conference on Computer Science and Software Engineering[C]. Wuhan,China, 2008. 1263-1266.
[6]	SILVIA T M , MARIKO N M , HECTOR P M . An image steganography systems based on BPCS and IWT[A]. 16th IEEE International Conference on Electronics,Communications and Computers[C]. Cholula, 2006. 51-51.
[7]	LEE K , JUNG C H , LEE S J ,et al. Color cube analysis for detection of LSB steganography in RGB color images[A]. ICCSA 2005[C]. LNCS 3481, 2005. 537-546.
[8]	KONG X W , LIU W F , YOU X G . Secret message location steganalysis based on local coherences of Hue[A]. PCM 2005[C]. LNCS 3768, 2005. 301-311.
[9]	陈铭, 张茹, 刘凡凡 . 基于区域相关性的LSB 匹配隐写分析[J]. 通信学报, 2010,31(3): 1-11. CHEN M , ZHANG R , LIU F F . Steganalysis of LSB matching based on regional correlation[J]. Journal on Communications, 2010,31(3): 1-11.
[10]	LYU S , FRAID H . Steganalysis using color wavelet statistics and one-class support vector machines[A]. Proceedings of SPIE,Electronic Imaging,Security,Steganalysis and Watermarking of Multimedia Contents[C]. VI,v5306, 2004. 35-45.
[11]	陈丹, 陈原, 王育民 . 一种针对彩色图象的泛盲掩密分析方案[J]. 电子与信息学报, 2005,27: 1542-1549. CHEN D , CHEN Y , WANG Y M . A universal blind steganalysis for color images[J]. Journal of Electronic and Information Technology, 2005,27: 1542-1549.
[12]	TU Y L , GONG S R . Universal steganalysis using color correlation and feature fusion[A]. 2008 International Symposium on Information Science and Engineering,Shanghai[C]. 2008. 107-111
[13]	毛家发, 林家骏, 戴蒙 . 基于图像攻击的隐藏信息盲检测技术[J]. 计算机学报, 2009,32(2): 318-327. MAO J F , LIN J J , DAI M . An attacked image based hidde essage blind detect technique[J]. Chinese Journal of Computer, 2009,32(2): 318-327
[14]	孙文颙, 刘婷婷, 张新鹏 ,等. 彩色图像通用隐写分析的多类统计特征[J]. 中国图象图形学报, 2008,13(10): 1914-1917. SUN W Y , LIU T T , ZHANG X P ,et al. Statistical features for universal steganalysis on color images[J]. Journal of Image and Graphics, 2008,13(10): 1914-1917.
[15]	MUMFORD D , SHAH J . Boundary detection by minimizing functions[A]. Proceedings of Conference on Computer Vision and Pattern Recognition[C]. San Francisco, 1985. 22-26.
[16]	MUMFORD D , SHAH J . Optimal approximation by piecewise smooth functions and associated variational problems[J]. Communication on Pure and Applied Mathematics, 1989,42(5): 577-685.
[17]	肖亮, 吴慧中, 韦志辉 . 面向彩色图像恢复与边缘检测的Mumford-Shah推广模型研究[J]. 计算机学报, 2006,29(2): 286-295. XIAO L , WU H Z , WEI Z H . Research on generalized mumford-shah model for color image restoration and edge detection[J]. Chinese Journal of Computers, 2006,29(2): 286-295.
[18]	PALUS H . Representations of color images in different spaces[A]. Sangwine S J,Horne R E N,eds.The Color Image Processing Handbook[C]. London:Chapman ＆Hall, 1998. 67-92.
[19]	HUANG J G , MUMFORD D . Statistics of natural images and models[A]. Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition[C]. Fort Collins, 1999. 1541-1547.
[20]	HUANG N E , SHEN Z , LONG S R ,et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings of the Royal Society of London,A, 1998,454: 903-995.
[21]	PHILIP G . Adding images to your site[EB/OL]. , 2010.
[22]	CHANG C C , LIN C J . LIBSVM:a library for support vector machines[EB/OL]. , 2001.

隐写方法（嵌入量）		Lyu^[10]			Chen^[11]			本文算法
隐写方法（嵌入量）	TP	FP	准确率	TP	FP	准确率	TP	FP	准确率
随机LSB（0.05bit/pixel）	59.6	36.7	61.5	70.8	35.5	67.7	77.1	23.2	77.0
随机LSB（0.1bit/pixel）	65.4	30.6	67.4	76.4	26.7	74.9	83.2	15.6	83.8
随机LSB（0.2bit/pixel）	81.8	25.2	78.3	86.7	19.1	83.8	92.1	9.5	91.3
随机LSB（0.4bit/pixel）	92.3	18.5	86.9	93.3	11.2	91.1	96.3	7.2	94.6
随机调制（0.05bit/pixel）	55.4	38.1	58.7	65.7	36.8	64.5	73.3	25.4	74.0
随机调制（0.1bit/pixel）	60.9	33.9	63.5	72.4	24.5	74.0	81.4	19.7	80.9
随机调制（0.2bit/pixel）	77.3	27.8	74.8	85.3	20.8	82.3	91.9	12.3	89.8
随机调制（0.4bit/pixel）	90.6	19.2	85.7	91.9	12.2	89.9	95.7	9.4	93.2
平均值	72.9	28.8	72.1	80.3	23.4	78.5	86.4	15.3	85.5

隐写方法	嵌入量		Lyu^[10]			Chen^[11]			本文算法
隐写方法	嵌入量	TP	FP	准确率	TP	FP	准确率	TP	FP	准确率
	10×10	58.7	40.5	59.1	64.1	36.9	63.6	71.2	25.6	72.8
Jsteg	20×20	61.2	31.6	64.8	70.4	32.2	69.1	77.4	17.8	79.8
	40×40	75.4	28.8	73.3	83.6	21.4	81.1	90.6	17.6	86.5
	80×80	87.1	9.7	88.7	91.8	9.4	91.2	94.9	7.3	93.8
	10×10	59.2	37.8	60.7	64.5	34.1	65.2	70.2	20.0	75.1
F5	20×20	65.2	32.6	66.3	76.1	26.5	74.8	81.7	17.1	82.3
	40×40	73.9	16.7	78.6	85.4	17.2	84.1	90.2	14.6	87.8
	80×80	86.2	13.0	86.6	91.9	10.3	90.8	94.1	7.3	93.4
	10×10	52.7	39.3	56.7	60.4	36.0	62.2	68.1	24.5	71.8
OutGuess0.2	20×20	66.4	28.0	69.2	76.5	27.1	74.7	83.7	20.3	81.7
	40×40	81.1	16.7	82.2	90.7	16.3	87.2	93.5	12.7	90.4
	80×80	U	U	U	U	U	U	U	U	U
	10×10	53.4	40.2	56.6	61.3	32.9	64.2	68.2	26.6	70.8
StegHide	20×20	61.6	32.8	64.4	70.9	25.5	72.7	76.5	22.1	77.2
	40×40	76.8	27.0	74.9	82.1	18.7	81.7	87.3	10.7	88.3
	80×80	86.3	11.7	87.3	93.5	9.1	92.2	94.9	7.7	93.6
	10×10	52.7	45.1	53.8	58.3	39.5	59.4	59.2	24.8	67.2
MB1	20×20	61.2	39.4	60.9	64.1	30.7	66.7	66.5	15.7	75.4
	40×40	77.5	32.5	72.5	80.4	23.2	78.6	83.5	21.3	81.1
	80×80	82.9	22.3	80.3	87.5	13.7	86.9	91.3	10.3	90.5
平均值		69.4	28.7	70.4	76.5	24.2	76.1	81.2	17.1	82.1

算法	特征向量	平均分类时间平均检测时间
Lyu^[10]	144	3.844s/幅8.457s/幅
Chen^[11]	49	1.789s/幅4.846s/幅
本文算法	4	0.335s/幅6.983s/幅