Using High-Dimensional Image Models to Perform Highly Undetectable Steganography

By Tomás Pevný, Tomáš Filler, Patrick Bas. Information Hiding, 2010

1,026 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/12453095157e2b096429526ed1e357fe00321333

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

Citations in the corpus

Deep Learning Hierarchical Representations for Image Steganalysis (2017, background): “domain include HUGO [1], WOW [2] and S-UNIWARD [3].”
Digital image steganography using universal distortion (2013, methodology): “Since both algorithms are highly adaptive, they are expected to better resists steganalysis using rich models [11] than HUGO.”
Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data (2020, methodology): “Previous techniques rely on Fourier transform [12, 8], JPEG compression [2, 1], or least significant bits modification [39, 22, 23].”
A new cost function for spatial image steganography (2014, methodology): “The cost function of HUGO (highly undetectable stego) [6] computes the weighted sum of difference between feature vectors extracted from a cover image and its stego version in SPAM (subtractive pixel adjacency matrix) [7] feature space.”
Rich Models for Steganalysis of Digital Images (2012, methodology): “, think of the second-order SPAM model with a large truncation threshold employed by HUGO [34]).”
Depth-Wise Separable Convolutions and Multi-Level Pooling for an Efficient Spatial CNN-Based Steganalysis (2020, neutral): “istics [1]–[3], [50], [51], [57].”
Automatic Steganographic Distortion Learning Using a Generative Adversarial Network (2017, methodology): “= (ρi,j )H×W is heuristically designed in a content-adaptive manner in all of the state-of-the-art spatial steganographic schemes, including HUGO [20], S-UNIWARD [21], MiPOD [22], and HILL [23].”
Hiding Images in Plain Sight: Deep Steganography (2017, methodology): “Advanced methods attempt to preserve the image statistics, by creating and matching models of the first and second order statistics of the set of possible cover images explicitly; one of the most popular is named HUGO [11].”
SteganoGAN: High Capacity Image Steganography with GANs (2019, methodology): “Given a set of N bits to be embedded, HUGO uses the distortion function to identify the top N pixels that can be modiﬁed while minimizing the total distortion across the image (Pevn´y et al., 2010).”
Selection-channel-aware rich model for Steganalysis of digital images (2014, background): “Examples include HUGO [17], WOW [13], and the UNIWARD family [14].”
HiDDeN: Hiding Data With Deep Networks (2018, methodology): “For example, some algorithms manipulate the least significant bits of some selected pixels [4]; others change mid-frequency components in the frequency domain [5].”
Generating steganographic images via adversarial training (2017, neutral): “HUGO [18] is considered to be one of the most secure steganographic techniques.”
Minimizing Additive Distortion in Steganography Using Syndrome-Trellis Codes (2011, methodology): “…embedding in steganography that minimizes a distortion function is connected to many basic principles used for constructing embedding schemes for complex cover sources today, including the principle of minimal-embedding-impact [16], approximate model-preservation [5], or the Gibbs construction [7].”
Universal distortion function for steganography in an arbitrary domain (2014, methodology): “1 Spatial domain In the spatial domain, we compare the proposed method with HUGO [15], HUGO implemented using the Gibbs construction with bounding distortion (HUGO BD) [4], WOW [17], LSBmatching (LSBM), and the edge-adaptive (EA) algorithm [28] With the exception of the EA algorithm, in which the co”
Ensemble Classifiers for Steganalysis of Digital Media (2012, background): “Because modern steganography [15], [35], [41] places embedding changes in those regions of images that are hard to model, increasingly more complex statistical descriptors of covers are required to capture a larger number of (weaker) dependencies among cover elements that might be disturbed by…”