Feature Extractors for Steganalysis

This webpage was created by DDE Lab at Binghamton University, NY, in December 2011 with the intention of providing the steganalysis research community with Matlab implementations of selected feature extractors for both JPEG and spatial domains.

DDE Lab keeps the copyright, however, the codes can be freely used for research and non-profit purposes. The full copyright notice is included in the header of all sourcecodes.

For suggestions and feedback, please use the contact information located at the bottom of this page.

Thank you.

Name Dim Download Domain Proposed Notes
Note: All JPEG domain extractors require Phil Sallee's MATLAB Jpeg Toolbox (jpeg_read routine). Update: the website is no longer active. With the author's permission, we provide the copy of relevant files (including a few pre-compiled verions) here.
CHEN 486 chen486.m JPEG [1] both inter- and intra-block Markov-based features
CC-CHEN 972 cchen972.m JPEG CHEN features enhanced by Cartesian calibration [3]
LIU** 216 liu216.m JPEG [4] all 216 features proposed in [4] are concatenated
CC-PEV 548 ccpev548.m JPEG [5,3] PEV features [5] enhanced by Cartesian calibration [3]*
SPAM 686 spam686.m Spatial [6] 2nd order Markov-based features
CDF 1,234 see Notes Both [7] union of CC-PEV and SPAM features
CC-C300 48,600 ccc300.m JPEG [8] our first high-dimensional rich model for JPEG steganalysis
CF* 7,850 cfstar.m JPEG [9] more compact rich model for JPEGs employing symmetrization
CC-JRM** 22,510 ccJRM.m JPEG [2] Cartesian Calibrated JPEG domain rich model, symmetrized, both integral and DCT-mode specific features
SRM 34,671 SRM.m
SRM.zip
SRM.tar
Spatial [10] full Spatial domain Rich Model (106 submodels). ZIP (for Windows) and TAR (for Linux) files contain C++ source code and Matlab MEX makefile. Extraction using the MEX file is much faster.
SRMQ1 12,753 SRMQ1.m
SRMQ1.zip
SRMQ1.tar
Spatial [10] Spatial domain Rich Model with the fixed quantization q=1c (see [10]). ZIP (for Windows) and TAR (for Linux) files contain C++ source code and Matlab MEX makefile. Extraction using the MEX file is much faster.
J+SRM 35,263 see Notes Both [2] union of SRMQ1 and CC-JRM
PSRM3
(PSRM8)
12870
(34320)
PSRM.m
PSRM.zip
PSRM.tar
Spatial [11] Projection Spatial Rich Model as published at SPIE 2013. The Matlab implementation is actually faster than its C++ (and MEX) Windows and Linux implementations
PSRM 12870 PSRM.m
PSRM.zip
PSRM.tar
Spatial [12] Projection Spatial Rich Model as published in TIFS. The Matlab implementation in the zip files is actually faster than its C++ (and MEX) Windows and Linux implementations
CSR 1183 CSR.m Spatial [13] Content-Selective Residuals as published at SPIE 2014. The attack is targeted at S-UNIWARD but can be easily modified.
DCTR 8000 DCTR.m
DCTR.zip
DCTR.tar
JPEG [14] Low complexity (FAST) features extracted from DCT residuals. ZIP (for Windows) and TAR (for Linux) files contain C++ source code and Matlab MEX makefile. Extraction using the MEX file is even faster.
maxSRM 34,671 (12,753) maxSRM.m
maxSRM.zip
maxSRM.tar maxSRMd2.m
maxSRMd2.zip
maxSRMd2.tar maxSRMq2d2.m
maxSRMq2d2.zip
maxSRMq2d2.tar
Spatial [15] Spatial domain Rich Model utilizing the approximate knowledge of the selection channel. The _d2 version uses a different shape (with respect to the traditional SRM) of the neighborhoods to select values of the residuals for the co-occurrence matrices. The _q2d2 further fixes the quantization q=2. ZIP and TAR files contain C++ source code and MEX files for Windows and Linux.
SCRMQ1, CRMQ1 12753 + 5404 SCRMQ1.m Spatial, color [16] Spatial and color Rich Model with the fixed quantization q=1. Dimension 18157 is for the default truncation Tc=3 in CRMQ1.
PHARM 12600 PHARM.m
PHARM.zip
PHARM.tar
JPEG [17] PHARM feature project in ZIP (for Windows) and TAR (for Linux) contains C++ source code and Matlab MEX makefile. Extraction using the MEX file much faster.

Important: Implementations for different systems (Matlab, Windows, Linux) will output different feature values due to their random generators. However, they should have identical performance.
CFA-aware CRM 5514, 4146, 10323 SRMQ1CFA.m Spatial, color [18] CFA-aware Color Rich Models with the fixed quantization q=1. More in readme.txt.
GFR 17000 GFR.m JPEG [19] JPEG Rich Model utilizing Gabor Filters.
sigma-features 1980 sigma-spamPSRM.m spatial [20] Selection-channel aware variant of the linear part of PSRM.
SCA-DCTR, SCA-GFR, SCA-PHARM various SCA_JPEG.zip JPEG [21] Selection-channel aware variant of various JPEG feature extractors.
PhaseAwareNet - PhaseAwareNet_SRC.zip JPEG [22] Caffe and MatConvNet implementation of JPEG-Phase-Aware Net.
SRNet - SRNet.zip Spatial, JPEG [23] TensorFlow implementation of SRNet.
JIN-SRNet - JIN_SRNet.zip Spatial, JPEG [24] Pytorch model of JIN pretrained SRNet.

* Note: The implementation of the CC-PEV features provided on this website is an updated version of our previously published implemenation available here. They differ in the DCT implementation. While the old CC-PEV implemenatation was dependent on the third party libraries DCT.c and IDCT.c, the new implementation (available on this webpage) uses implicit DCT incorporated in Matlab's imwrite function, which makes it more user-friendly. Both implementations can be used and they should give similar results. However, they cannot be used interchangeably! For example, if cover images were processed using the old extractor and stego images using the new one, detection errors would be artificially decreased as we would be detecting not only embedding impact but also differences in the JPEG compressor (DCT implementation).

** July 2014: Corrected a mistake discovered by Yi Zhang, feature subsets Ax_T5 and Ax_T5_ref in cc-JRM set and liu_absNJ_2_c in LIU set were always zeros due to use of incorrect function.

Contact

  • Jessica Fridrich - fridrich (at) binghamton (dot) edu
  • Vojtech Holub - vojtech_holub (at) yahoo (dot) com
  • Tomas Denemark - tdenema1 (at) binghamton (dot) edu

References

[1] C. Chen, Y. Q. Shi, JPEG image steganalysis utilizing both intrablock and interblock correlations, IEEE ISCAS, International Symposium on Circuits and Systems, pages 3029–3032, May 2008.

[2] J. Kodovsky, J. Fridrich, Steganalysis of JPEG Images Using Rich Models, Proc. SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics XIV, San Francisco, CA, January 23–25, 2012. [pdf] [slides]

[3] J. Kodovsky, J. Fridrich, Calibration revisited, In J. Dittmann,S. Craver, and J. Fridrich, editors, Proceedings of the 11th ACM Multimedia and Security Workshop, Princeton, NJ, September 7–8, 2009. [ pdf ] [ slides ]

[4] Q. Liu, Steganalysis of DCT–embedding based adaptive steganography and YASS, In J. Dittmann, S. Craver, and C. Heitzenrater, editors, Proceedings of the 13th ACM Multimedia & Security Workshop, pages 77–86, Niagara Falls, NY, September 29–30, 2011.

[5] T. Pevny and J. Fridrich, Merging Markov and DCT features for multiclass JPEG steganalysis, In E. J. Delp and P. W. Wong, editors, Proceedings SPIE, Electronic Imaging, Security, Steganography, and Watermarking of Multimedia Contents IX, volume 6505, pages 3 1–3 14, San Jose, CA, January 29–February 1, 2007. [ pdf ]

[6] T. Pevny, P. Bas, and J. Fridrich, Steganalysis by Subtractive Pixel Adjacency Matrix IEEE Trans. on Info. Forensics and Security, vol. 5(2), pp. 215�224, 2010. [ pdf ]

[7] J. Kodovsky, T. Pevny, and J. Fridrich, Modern Steganalysis Can Detect YASS, Proc. SPIE, Electronic Imaging, Media Forensics and Security XII, pages 2 1–2 11, San Jose, CA, January 17–21, 2010. [pdf] [slides]

[8] J. Kodovsky and J. Fridrich, Steganalysis in high dimensions: fusing classifiers built on random subspaces, Proc. SPIE, Electronic Imaging, Media, Watermarking, Security and Forensics XIII, San Francisco, CA, January 23–26, 2011. [pdf] [slides]

[9] J. Kodovsky, J. Fridrich, and V. Holub, Ensemble classifiers for steganalysis of digital media, IEEE Transactions on Information Forensics and Security, 2012. [pdf] [download section]

[10] J. Fridrich and J. Kodovsky, Rich models for steganalysis of digital images, IEEE Transactions on Information Forensics and Security. [pdf]

[11] V. Holub, J. Fridrich and T. Denemark, Random Projections of Residuals as an Alternative to Co-occurrences in Steganalysis, Proc. SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics XV, vol. 8665, San Francisco, CA, February 3–7, 2013. [pdf] [slides]

[12] V. Holub and J. Fridrich, Random Projections of Residuals for Digital Image Steganalysis, IEEE Transactions on Information Forensics and Security, vol. 8, no. 12, pp. 1996–2006, December 2013. [pdf]

[13] T. Denemark, J. Fridrich and V. Holub, Further study on security of S–UNIWARD, SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics, vol. 9028, San Francisco, CA, February 2–6, 2014. [pdf]

[14] V. Holub and J. Fridrich, Low Complexity Features for JPEG Steganalysis Using Undecimated DCT, IEEE Transactions on Information Forensics and Security, to appear. [pdf]

[15] T. Denemark, V. Sedighi, V. Holub, R. Cogranne and J. Fridrich, Selection-Channel-Aware Rich Model for Steganalysis of Digital Images, IEEE Workshop on Information Forensic and Security, Atlanta, GA, December 3–5, 2014. [pdf]

[16] M. Goljan, J. Fridrich, and R. Cogranne, Rich Model for Steganalysis of Color Images, IEEE Workshop on Information Forensic and Security, Atlanta, GA, December 3–5, 2014. [pdf]

[17] V. Holub and J. Fridrich, Phase-Aware Projection Model for Steganalysis of JPEG Images, Proc. SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics XVII, vol. 9409, San Francisco, CA, February 8–12, 2015. [pdf]

[18] M. Goljan and J. Fridrich, CFA-aware Features for Steganalysis of Color Images, Proc. SPIE, Electronic Imaging, Media Watermarking, Security, and Forensics XVII, to appear, San Francisco, CA, February 8–12, 2015.

[19] X. Song, F. Liu, C. Yang, X. Luo and Y. Zhang, Steganalysis of Adaptive JPEG Steganography Using 2D Gabor Filters, Proceedings of the 3rd ACM Workshop on Information Hiding and Multimedia Security. ACM, 2015.

[20] T. Denemark, J. Fridrich and P. Comesana Alfaro, Improving Selection-Channel-Aware Steganalysis Features, IS&T International Symposium on Electronic Imaging 2016.

[21] T. Denemark, M. Boroumand, J. Fridrich, Steganalysis Features for Content-Adaptive JPEG Steganography, IEEE TIFS, vol. 11, no. 8, pp. 1736-1746, August 2016.

[22] Mo Chen, Vahid Sedighi, Mehdi Boroumand, Jessica Fridrich, JPEG-Phase-Aware Convolutional Neural Network for Steganalysis of JPEG Images, Proceedings of IH&MMSec 17, Philadelphia, PA, USA, June 20-22, 2017.

[23] Mehdi Boroumand, Mo Chen, Jessica Fridrich, Deep Residual Network for Steganalysis of Digital Images, IEEE TIFS, to appear.

[24] Jan Butora, Yassine Yousfi, Jessica Fridrich, How to Pretrain for Steganalysis, 9th IH&MMSec. Workshop, Brussels, Belgium, June 22-25, 2021. [pdf]


Last update: Oct. 2018