Steganalysis using Regional Correlation and Second-order Markov Features

다국어 초록 (Multilingual Abstract)

A blind steganalysis based on regional correlation and second-order Markov transition probability matrix is proposed for JPEG images. By analyzing the region correlation of JPEG image, the Markov transition probability matrix is used to capture the correlation of intra-block and inter-block DCT coefficients. In addition, the calibrated features are extracted from the calibrated images. The difference features between the original and calibrated images are used for training and classifying. Three different image libraries are used to detect the five kinds of typical JPEG steganography schemes. Experimental results show that, in comparison to the several effective steganalysis, the proposed scheme improves the detection accuracy on some JPEG-based steganography schemes, including outguess, Steghide and MB1.

목차 (Table of Contents)

• Abstract
• 1. Introduction
• 2. Feature Generation
• 2.1. Image Scanning Patterns
• 2.2. Intra-block Features
• 2.3. Inter-block Features
• 2.4. Image Calibration
• 3. Experimental Results
• 3.1. Image Database and JPEG Steganography Tools
• 3.2. Experimental Results
• 4. Conclusion
• Acknowledgements
• References