Abstract
Calibration attack is one of the most important attacks specially on JPEG domain steganography. Prediction of cover image statistics from its stego image is an important requirement in calibration based attacks. Domain separation techniques are used as a counter measures against such attacks because they make the cover image prediction process rather difficult. Most of the algorithms in the past are based on randomization of embedding locations. In this paper, we propose a new domain separation technique which is based on randomization of DCT kernel matrix. A comprehensive set of experiments are done to validate that the proposed domain separation scheme performs better than the related existing schemes.
Similar content being viewed by others
Notes
All the source codes for the feature extraction except 274DCA and ensemble classifier are taken from the link dde.binghamton.edu website
References
“Further study on yass: Steganography based on randomized embedding to resist blind steganalysis,” in Proc. SPIE - Security, Steganography, and Watermarking of Multimedia Contents X, vol. 6819, Jan. 2008, pp. 681 917–681 928.
“Steganalysis in high dimensions: fusing classifiers built on random subspaces,” in Proc. SPIE, vol. 7880, 2011.
Avcibas I, Memon N, Sankur B (2003) Steganalysis using image quality metrics. IEEE Transaction on Image Processing 12:221–229
Chandrakanth B, Dutta T, Sur A “Steganographic algorithm based on parametric discrete cosine transform,” in Proc. ICVGIP’12 Eighth Indian Conference on Computer Vision, Graphics and Image Processing Article No. 79.
Chen B, Wornell G (2001) Quantization index modulation: a class of provably good methods for digital watermarking and information embedding. IEEE Trans Inf Theory 47:1423–1443
Eggers J, Bauml R, Girod B “A communications approach to image steganography,” in Proc. SPIE Security and Watermarking of Multimedia Contents IV, vol. 4675, Apr. 2002.
Farid H, Siwei L “Detecting hidden messages using higher-order statistics and support vector machines,” in Proc. 5th International Workshop on Information Hiding, Oct. 2002, pp. 340–354.
Fridrich J “Feature-based steganalysis for jpeg images and its implications for future design of steganographic schemes,” in Proc. 6th International Workshop on Information Hiding, May 2004, pp. 67–81.
Fridrich J, Goljan M, Lison P, Soukal D (2005) “Writing on wet paper”, IEEE Transactions on Signal Processing. Special Issue on Media Security 53:3923–3935
Holub V, Fridrich J, Denemark T “Universal distortion function for steganography in an arbitrary domain,” EURASIP Journal on Information Security, 2014
Jain AK Fundamentals of Digital Image Processing. Prentice Hall, 1989, pp. 150–153.
Kodovsky J, Fridrich J “Calibration revisited,” in Proc. SPIE, 11th ACM Multimedia and Security Workshop, p. 313.
Kodovsky J, Fridrich J “Statistically undetectable jpeg steganography: Dead ends, challenges, and opportunities,” in Proc. ACM Multimedia and Security Workshop, 2007, pp. 3–14.
Kodovsky J, Fridrich J, Holub V (2012) Ensemble classifiers for steganalysis of digital media. IEEE Transactions on Information Forensics and Security 7:432–444
Kodovsky J, Pevny T, Fridrich J “Modern steganalysis can detect yass,” in Proc. SPIE, Electronic Imaging, Media Forensics and Security XII, vol. 7541, 2010.
Liu Q “Steganalysis of dct-embedding based adaptive steganography and yass,” in Proc. 13th ACM Multimedia Workshop on Multimedia and Security, p. 77.
Pevny T, Bas P, Fridrich J “Steganalysis by subtractive pixel adjacency matrix,” IEEE Trans. on Info. Forensics and Security, vol. 5, p. 215.
Pevny T, Fridrich J “Merging markov and dct features for multi-class jpeg steganalysis,” in Proc. SPIE, Electronic Imaging, Security, Steganography, and Watermarking of Multimedia Contents IX, vol. 6505, Jan. 2007, pp. 03–04
Pevny T, Fridrich J “Merging markov and dct features for multiclass jpeg steganalysis,” in Proc. SPIE, Electronic Imaging, Security, Steganography, and Watermarking of Multimedia Contents IX, p. 313.
Provos N “Defending against statistical steganalysis,” in Proc. 10th USENIX Security Symposium, vol. 10, 2001, pp. 24–24.
Sallee P “Model-based steganography,” in Proc. 2nd International Workshop on Digital Watermarking, Oct. 2003, pp. 154–167.
Solanki K, Sarkar A, Manjunath B “Yass: Yet another steganographic scheme that resists blind steganalysis,” in Proc. 9th International Workshop on Information Hiding, Jun. 2007, pp. 16–31.
Solanki K, Sullivan K, Madhow U, Manjunath B, Chandrasekaran S (2005) “Statistical restoration for robust and secure steganography”, in Proc. IEEE Int Conf on Image Processing 2:1118–1121
Sur A, Vignesh R, Mukherjee J “Secure steganography using randomized cropping,” LNCS Transactions on Data Hiding and Multimedia Security, Springer, vol. 7110, pp. 82–95, Feb. 2012
Upham D, “Steganographic algorithm jsteg.”
Wang JLJ, Wiederhold G (2001) Simplicity : semantics-sensitive integrated matching for picture libraries. IEEE Trans on Pattern Analysis and Machine Intelligence 23:947–963
Westfeld A “High capacity despite better steganalysis (f5 - a steganographic algorithm),” in Proc. 4th Int. Workshop on Information Hiding, Apr. 2001, pp. 289–302
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix I: example of PDCT
Consider a matrix A
The 3 × 3 DCT kernel is given by D
DCT(A) = D ∗ A ∗ D′
The 3 × 3 PDCT kernel when α = 1.25 is given by \( \overline{D} \)
PDCT(A) = \( \overline{D}\ast A\ast {\overline{D}}^{\prime } \)
Appendix II: Reversibility of randomized DCT (RDCT)
Proof for \( {\overline{D}}^{\prime}\overline{D}= I \), where \( \overline{D} \) is the RDCT kernel matrix
Let D 1, D 2, D 3, D 4, D 5, D 6, D 7, D 8 be the column vectors of DCT matrix D. We know that the DCT matrix D is a orthonormal matrix where the columns are unit vectors perpendicular to each other.
i.e. D i . D i = 1 and D i . D j = 0. This gives D′D = I.
Now consider the product \( {\overline{D}}^{\prime}\overline{D} \). The diagonal entries are of the form D i . D i and other entries are of the form D i . D j . Hence it is clear that \( {\overline{D}}^{\prime}\overline{D}= I \)
Proof for Inverserdct(A rdct ) = A
From Eq. 3 we get,
As we have proved \( {\overline{D}}^{\prime}\overline{D}= I \), the above reduces to A. Therefore
Hence the reversibility is proved.
Appendix III: reversibility example of RDCT transform matrix
Consider a matrix A
The 3 × 3 DCT kernel is given by D
Let the random matrix K
The 3 × 3 RDCT kernel is given by \( \overline{D} \) = KD
RDCT(A) = A rdct = \( \overline{D}\ast A\ast {\overline{D}}^{\prime } \)
\( Inverserdct\left({A}_{rdct}\right)={\overline{D}}^{\prime }{A}_{rdct}\overline{D} \)
Rights and permissions
About this article
Cite this article
Karri, S., Sur, A. Steganographic algorithm based on randomization of DCT kernel. Multimed Tools Appl 74, 9207–9230 (2015). https://doi.org/10.1007/s11042-014-2077-0
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11042-014-2077-0