EFFECTIVENESS OF STEGANALYSIS BASED ON MACHINE LEARNING METHODS

Sivachev Aleksey V., Prokhozhev Nikolai N, Mikhailichenko Olga V, Bashmakov Daniil A.

doi:10.17586/2226-1494-2017-17-3-457-466

2017 , VOLUME 17, NUMBER 3 ( May-June )

ISSN 2226-1494 (print), ISSN 2500-0373 (online)

Publications

Editor-in-Chief

Nikiforov
Vladimir O.
D.Sc., Prof.

Partners

doi: 10.17586/2226-1494-2017-17-3-457-466

EFFECTIVENESS OF STEGANALYSIS BASED ON MACHINE LEARNING METHODS

A. V. Sivachev, N. N. Prokhozhev, O. V. Mikhailichenko, D. A. Bashmakov

Read the full article

Article in Russian

For citation: Sivachev A.М., Prokhozhev N.N., Mikhailichenko O.V., Bashmakov D.A. Effectiveness of steganalysis based on machine learning methods. Scientific and Technical Journal of Information Technologies, Mechanics and Optics, 2017, vol. 17, no. 3, pp. 457–466 (in Russian). doi: 10.17586/2226-1494-2017-17-3-457-466

Abstract

Subject of Study.The paper presents comparative accuracy estimation of modern machine learning-based steganalytic methods. The paper deals with the most perspective methods in tasks of the passive counteraction to the information transfer channels using the discrete wavelet domain of static digital images. Methods. We have studied methods proposed by Gireesh Kumar, Hany Farid, Changxin Liu, Yun Q. Shi and the SPAM method. Basically the methods apply statistical moments obtained from wavelet bands LL, HL, LH and HH, as well as additional image features forming a support vector. BOWS2 image collection was used to estimate the effectiveness of methods. Steganographic impact was modeled by changing the least significant bits of coefficients for the each DWT band (LL, LH, HL and HH) with 5% and 20% of payload. The effectiveness of explored methods is estimated in view of obtained true positive, true negative, false positive and false negative image classification values. Main Results. The study has shown that all explored methods except for SPAM are effective in the task of detecting of embedding in HH band. As for the detection of the embedding in LH band, Yun Q. Shi is the most effective algorithm. In the task of the detecting in HL band, all explored methods except for SPAM have appeared to be comparatively effective under condition of big payload. When detecting the embedding in LL band, all methods have shown the effectiveness about 50% regardless the payload rate. It is established that the considered methods are not able to render effective counteraction to the hidden data channel, using the LH and HL region due to the fact that they use Haar wavelet transform. It is concluded that the application of the optimal wavelet transform makes it possible to reduce the intersection area of value histograms of the first statistical moment for the original images and steganoimages. Practical Relevance. The work results are useful to the specialists in the field of information security in the tasks of detection and combating the hidden data channels. The obtained results can be used in the development of steganalysis systems and improved methods of steganalysis as well.

Keywords: steganography, machine learning, passive counteraction, hidden channel, steganalisys system and algorithms, binary classification, one-dimensional DWT low-frequency region, discrete wavelet transform, Haar and Daubechies transform, least significant bit

References

1. Gribunin V.G., Okov I.N., Turintsev I.V. Digital Steganography. Moscow, SOLON-Press, 2016, 262 p. (In Russian)

2. Makarenko S. The steganographic system interconnection basic reference model and the justification of new areas of steganography theory's development. Voprosy Kiberbezopasnosti, 2014, no. 2, pp. 24–32.(In Russian)

3. Steganography: A Powerful Tool for Terrorists and Corporate Spies. Stratfor. Available at: https://www.stratfor.com/analysis/steganography-powerful-tool-terrorists-and-corporate-spies (accessed 02.02.2017).

4. Kessler G.C. An overview of steganography for the computer forensics examiner. Forensic Science Communications, 2004, vol. 6, no. 3.

5. Gayathri C., Kalpana V. Study on image steganography techniques. International Journal of Engineering and Technology, 2013, vol. 5, no.2, pp.572–577.

6. Prokhozhev N., Mikhailichenko O., Sivachev A., Bashmakov D., Korobeynikov A.G. Passive steganalysis evaluation: reliabilities of modern quantitative steganalysis algorithms. Advances in Intelligent Systems and Computing, 2016,vol. 451, pp. 89–94. doi:10.1007/978-3-319-33816-3_9

7. Prokhozhev N.N., Mikhailichenko O.V., Bashmakov D.A., Sivachev A.V., Korobeinikov A.G. Software Study the effectiveness of statistical algorithms of quantitative steganalysis in the task of detecting hidden information channels. Systems and Computational Methods, 2015, no. 3, pp. 281–292. (In Russian) doi: 10.7256/2305-6061.2015.3.17233

8. Schaathun H.G.Machine Learning in Image Steganalysis. Wiley-IEEE Press, 2012, 290 p.

9. GireeshKumarT., JithinR., Shankar D.D. Featurebasedsteganalysis using wavelet decomposition and magnitude statistics. Proc. Int. Conf. on Advances in Computer Engineering. Bangalore, India, 2010, pp. 298–300. doi: 10.1109/ACE.2010.33

10. Hany F. Detecting Steganographic Messages in Digital Images. Technical Report TR2001-412, Dartmouth College, 2001.

11. Liu C., Ouyang C., Guo M., Chen H. Image steganalysis based on spatial domain and DWT domain features.Proc. 2^nd int. Conf. on Networks Security, Wireless Communications and Trusted Computing, 2010,vol.1, pp. 329–331. doi: 10.1109/NSWCTC.2010.271

12. Pevny T., Bas P., Fredrich J. Steganalysis by subtractive pixel adjacency matrix. IEEE Transactions on Information Forensics and Security, 2010, vol. 5, no. 2, pp. 215–224. doi: 10.1109/TIFS.2010.2045842

13. Shi Y.Q., XuanG., Yang C., Gao J., Zhang Z., Chai P., Zou D., Chen C., Chen W. Effective steganalysis based on statistical moments of wavelet characteristic function. Proc. Int. Conf. on Information Technology: Coding and Computing, ITCC'05. Las Vegas, USA, 2005, vol. 2, pp. 768–773.

14. BOWS2 the 10000 original images. Available at: http://bows2.ec-lille.fr/ (accessed 12.04.2017).

15. WaliaR. Steganography based on neighborhood pixels. Proc. 2^nd Conf. on Advances in Computing, Communications and Informatics, ICACCI. Mysore, India,2013, pp. 203–206. doi: 10.1109/ICACCI.2013.6637171

16. Qin J., Xiang X., Deng Y., Li Y., Pan L. Steganalysis of highly undetectable steganography using convolution filtering. Information Technology Journal, 2014, vol. 13, no. 16, pp. 2588–2592.doi: 10.3923/itj.2014.2588.2592

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License