An approach to increase the JPHIDE steganograms detection accuracy

Fìz.-mat. model. ìnf. tehnol. 2021, 32:170-174

Abstract

The paper proposes a method for improving the accuracy of steganoanalytical systems that use an ensemble classifier. The method involves a weighted final vote of several highly sensitive models of characteristic vectors. Its effectiveness was evaluated for the task of detecting steganograms created by the Jphide program. The accuracy obtained by usage of one of the models: LIU, CC-PEV, CC-C300, DCTR, PHARM, GFR and with using a combination of several models according to the developed method was compared. The test results proved that the weighted final voting of several highly sensitive models does increase the accuracy of the detection of steganograms with a relatively small payload (short secret messages) without compromising the accuracy of the detection of steganograms with a high payload.

References

1. Koshkina, N. V. (2020). Comparison of Efficiency of Statistical Models Used for Formation of Feature Vectors by JPEG Images Steganalysis. Theoretical and Applied Cybersecurity, 2(1), 22-28.
   DOI doi.org/10.20535/tacs.2664-29132020.1.209433
2. Koshkina, N. V. (2020). Research of Main Components of Machine Learning Based JPEG-Steganalysis Systems. Ukrainian Information Security Research Journal, 22(2), 97-108. http://jrnl.nau.edu.ua/index.php/ZI/article/view/14801/21490
3. Holub, V., Fridrich, J. (2015). Phase-Aware Projection Model for Steganalysis of JPEG Images, Proc. SPIE. Electronic Imaging, Media Watermarking, Security, and Forensics XVII.
   DOI doi.org/10.1117/12.2075239