Merging Markov and DCT features for multi-class JPEG steganalysis

By Tomás Pevný, Jessica J. Fridrich. Electronic imaging, 2007

567 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/163cfb244dce5c0c050a56240b37b3689b58c2d4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Jessica J. Fridrich
2. Vojtech Holub
3. Bin Li
4. Jan Kodovský
5. Jiwu Huang
6. Tomás Pevný
7. Gongshen Liu
8. Feng Zhu

Citations in the corpus

• Digital image steganography: Survey and analysis of current methods (2010, methodology): “In the frequency domain, Pevny and Jessica [113] developed a multi-class JPEG steganalysis system that comprised of DCT features and calibrated Markov features, which were then merged to produce a 274-dimensional feature vector.”
• Benchmarking for Steganography (2008, methodology): “In this paper, we selected the 274-dimensional feature vector described in [32].”
• Steganalysis of Adaptive JPEG Steganography Using 2D Gabor Filters (2015, methodology): “The features extraction by subsamp-ing the filtered image into 64 subimages according to the 64 DCT modes in 8 × 8 DCT block can enhance the diversity and effectiveness of the steganalysis features, the similar idea has been used for the local histogram features extraction in literature [20, 14].”
• Steganalysis of JPEG images using rich models (2012, methodology): “• CHEN (486) = Markov features utilizing both intra- and inter-block dependencies [1], • CC-CHEN (972) = CHEN features improved by Cartesian calibration [8], • LIU (216) = the union of diff-absNJ-ratio and ref-diff-absNJ features published in [14], • CC-PEV (548) = Cartesian-calibrated PEV feature s”
• Statistically undetectable jpeg steganography: dead ends challenges, and opportunities (2007, methodology): “We select the 274-dimensional feature vector consisting of 193 extended DCT features and 81 Markov features described in [12, 33, 29].”
• Depth-Wise Separable Convolutions and Multi-Level Pooling for an Efficient Spatial CNN-Based Steganalysis (2020, methodology): “The commonly used statistical features include the binary similarity measure feature [7], DCT [8], [9] and wavelet coefficient feature [10], co-occurrence matrix feature [11] and so on.”
• Steganalysis by Subtractive Pixel Adjacency Matrix (2009, methodology): “Prior to all experiments, all databases were divided into training and testing subsets with approximately the same number of images.”
• An improved approach to steganalysis of JPEG images (2010, methodology): “By applying calibration to Markov features, Pevny and Fridrich merged the DCT features and calibrated Markov features to improve the steganalysis performance in JPEG images [23].”
• A Survey on Image Steganography and Steganalysis (2011, methodology): “[34, 35, 38, 39] can reliably detect many JPEG steganographic schemes but may not be effective to spatial steganography.”
• Ensemble Classifiers for Steganalysis of Digital Media (2012, methodology): “The feature set designed for JPEG images described in [42] used 274 features and was later extended to twice its size [28] by Cartesian calibration, while 324-and 486-dimensional feature vectors were proposed in [48] and [11], respectively.”