Steganalysis Features for Content-Adaptive JPEG Steganography

Denemark, Tomáš; Boroumand, Mehdi; Fridrich, Jessica

doi:10.1109/tifs.2016.2555281

Cited by 164 publications

(88 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…J-UNIWARD served as the only steganographic method in this study. According to the steganalysis results presented in [8,9,11], J-UNIWARD should be the most secure algorithm embedding in the JPEG domain. Embedding rates of 0.1, 0.2, 0.3, and 0.4 bpnzAC were selected for experiments.…”

Section: Methodsmentioning

confidence: 99%

“…Similar to what has been done in the traditional feature-based steganalysis methods [8,9,11], the input in the JPEG format are first transformed to the spatial domain (without the last rounding step), then go through a set of filter banks. In the proposed CNN, undecimated DCT of size 4×4 are selected to project every single input to 16 different frequency bands.…”

Section: The Proposed Cnn Achitecturementioning

confidence: 99%

“…Other types of filter banks, e.g., the Garbor filters [9], have not been studied yet and should belong to future research. Same as in [7][8][9]11], in this work, only the magnitudes of the subbands are used, and they are further truncated with a slightly tuned global threshold value of 8. Quantization as another essential ingredient in the traditional feature extraction procedure has been abandoned to prevent unnecessary information loss; after all, the CNNs do not explicitly assemble histograms for statistical modeling.…”

Section: The Proposed Cnn Achitecturementioning

confidence: 99%

See 2 more Smart Citations

Deep Convolutional Neural Network to Detect J-UNIWARD

Xu¹

2017

Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security

203

154

View full text Add to dashboard Cite

This paper presents an empirical study on applying convolutional neural networks (CNNs) to detecting J-UNIWARDone of the most secure JPEG steganographic method. Experiments guiding the architectural design of the CNNs have been conducted on the JPEG compressed BOSSBase containing 10,000 covers of size 512×512. Results have verified that both the pooling method and the depth of the CNNs are critical for performance. Results have also proved that a 20-layer CNN, in general, outperforms the most sophisticated feature-based methods, but its advantage gradually diminishes on hard-to-detect cases. To show that the performance generalizes to large-scale databases and to different cover sizes, one experiment has been conducted on the CLS-LOC dataset of ImageNet containing more than one million covers cropped to unified size of 256×256. The proposed 20-layer CNN has cut the error achieved by a CNN recently proposed for large-scale JPEG steganalysis by 35%. Source code is available via GitHub: https://github.com/GuanshuoXu/deep_cnn_jpeg_steganalysis

show abstract

Section: Methodsmentioning

confidence: 99%

Section: The Proposed Cnn Achitecturementioning

confidence: 99%

Section: The Proposed Cnn Achitecturementioning

confidence: 99%

See 1 more Smart Citation

Deep Convolutional Neural Network to Detect J-UNIWARD

Xu¹

2017

Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security

203

154

View full text Add to dashboard Cite

show abstract

“…DCTR. SCA proposed in [18] is a selection-channel-aware variant of JPEG rich models targeted at content-adaptive JPEG steganography 1 .…”

Section: Introductionmentioning

confidence: 99%

Large-Scale JPEG Image Steganalysis Using Hybrid Deep-Learning Framework

Zeng

Tan

et al. 2018

IEEE Trans.Inform.Forensic Secur.

190

View full text Add to dashboard Cite

Abstract-Adoption of deep learning in image steganalysis is still in its initial stage. In this paper we propose a generic hybrid deep-learning framework for JPEG steganalysis incorporating the domain knowledge behind rich steganalytic models. Our proposed framework involves two main stages. The first stage is hand-crafted, corresponding to the convolution phase and the quantization & truncation phase of the rich models. The second stage is a compound deep neural network containing multiple deep subnets in which the model parameters are learned in the training procedure. We provided experimental evidences and theoretical reflections to argue that the introduction of threshold quantizers, though disable the gradient-descent-based learning of the bottom convolution phase, is indeed cost-effective. We have conducted extensive experiments on a large-scale dataset extracted from ImageNet. The primary dataset used in our experiments contains 500,000 cover images, while our largest dataset contains five million cover images. Our experiments show that the integration of quantization and truncation into deeplearning steganalyzers do boost the detection performance by a clear margin. Furthermore, we demonstrate that our framework is insensitive to JPEG blocking artifact alterations, and the learned model can be easily transferred to a different attacking target and even a different dataset. These properties are of critical importance in practical applications.

show abstract

“…According to Hopper [25] and Denemark [26] the steganography involves the use of certain terminologies, such as cover or host, embedding, stego-object, and extraction. Figure 5 discussed types of steganography terminologies.…”

Section: Steganography Terminologiesmentioning

confidence: 99%

A Survey of Spatial Domain Techniques in Image Steganography

Awad

2018

eduj

View full text Add to dashboard Cite

This paper presents the most techniques of spatial domain that used in image steganography where it is used to hide confidential message under a cover image. Steganography is a much better way to convey a secret message compared to cryptography. This is because cryptography is outlawed while steganography can avoid such policies to pass message covertly. Steganography is one of the most common techniques of securing information, which insists on concealing the clandestine data within the cover image in a way that nobody suspects to its existence. The challenge of steganography techniques is to make a reasonable balance between the file and the size of data that can be exchanged. Furthermore, the strength of the strategy and the security of the mysterious data are the fact that can't be hidden.

show abstract

Steganalysis Features for Content-Adaptive JPEG Steganography

Cited by 164 publications

References 22 publications

Deep Convolutional Neural Network to Detect J-UNIWARD

Deep Convolutional Neural Network to Detect J-UNIWARD

Large-Scale JPEG Image Steganalysis Using Hybrid Deep-Learning Framework

A Survey of Spatial Domain Techniques in Image Steganography

Contact Info

Product

Resources

About