Analysis of the Scalability of a Deep-Learning Network for Steganography “Into the Wild”

Ruiz, Hugo; Chaumont, Marc; Yedroudj, Mehdi; Amara, Ahmed Oulad; Comby, Frédéric; Subsol, Gérard

doi:10.1007/978-3-030-68780-9_36

Cited by 11 publications

(6 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We do not see much of a difference in performance between IN and JIN for both SRNet and B0. We think this might be due to fact that the training set is sufficiently large and the network architecture cannot get better anymore, a behavior that was already observed in [23]. Nevertheless, one interesting observation remains.…”

Section: Jpeg Domainmentioning

confidence: 86%

How to Pretrain for Steganalysis

Butora

Yousfi

Fridrich

2021

Proceedings of the 2021 ACM Workshop on Information Hiding and Multimedia Security

View full text Add to dashboard Cite

In this paper, we investigate the effect of pretraining CNNs on Ima-geNet on their performance when refined for steganalysis of digital images. In many cases, it seems that just 'seeing' a large number of images helps with the convergence of the network during the refinement no matter what the pretraining task is. To achieve the best performance, the pretraining task should be related to steganalysis, even if it is done on a completely mismatched cover and stego datasets. Furthermore, the pretraining does not need to be carried out for very long and can be done with limited computational resources. An additional advantage of the pretraining is that it is done on color images and can later be applied for steganalysis of color and grayscale images while still having on-par or better performance than detectors trained specifically for a given source. The refining process is also much faster than training the network from scratch. The most surprising part of the paper is that networks pretrained on JPEG images are a good starting point for spatial domain steganalysis as well.

show abstract

Section: Jpeg Domainmentioning

confidence: 86%

How to Pretrain for Steganalysis

Butora

Yousfi

Fridrich

2021

Proceedings of the 2021 ACM Workshop on Information Hiding and Multimedia Security

View full text Add to dashboard Cite

show abstract

“…One of the most important advantages of this method is the reliability of more than one type of image, but this method suffers from its inability to account for hacker attacks. The DLbased SS [24,25] was implemented to improve the NN classifier, which achieves improved security and data hiding capacity in the network. The information hiding method is good and effective, and it cannot embed a large amount of data, which most traditional methods suffer from.…”

Section: Related Workmentioning

confidence: 99%

Improved Security of a Deep Learning-Based Steganography System with Imperceptibility Preservation

Fadhil

Jalo

Mohammad

2023

Int. j. electr. comput. eng. syst. (Online)

View full text Add to dashboard Cite

Since its inception, the steganography system (SS) has continuously evolved and is routinely used for concealing various sensitive data in an imperceptible manner. To attain high performance and a better hiding capacity of the traditional SS, it has become essential to integrate them with diverse modern algorithms, especially those related to artificial intelligence (AI) and deep learning (DL). Based on this fact, we proposed a DL-based SS (DLSS) to extract some significant features (like pixel locations, importance, and proximity to the imperceptibility) from the cover image using a neural network (NN) in a hierarchical form, thus selecting the candidate pixels for embedding afterwards. The pixel weight was expressed in terms of the position, imperceptibility, and its relationship with adjacent pixels to be a stego image. Performance evaluation revealed that the proposed DLSS achieved imperceptibility of 84 dB for images in training mode of a standard dataset.

show abstract

“…We want to generate (from the RAW database) many learning datasets of different sizes from ten thousand to two million grey-scale JPEG images. One possible use is for evaluating the scalability of a steganalysis network, as in [15]. It is also necessary to set up a test dataset that will be the same for all learning datasets.…”

Section: Training Database Constructionmentioning

confidence: 99%

“…This database was used to make a test on scalability of a network in [15]. In this work, we applied the algorithm J-UNIWARD developed by Holub et al [11] with a payload of 0.2 bpnzacs (bits per non zero AC coefficients).…”

Section: Format Of Imagesmentioning

confidence: 99%

“…Moreover, GPU computation capabilities increase regularly, which ensures faster computing speed which reduces Nevertheless, in most cases, the performance of steganalysis based on Deep Learning depends on the size of the learning image database. To a certain extent, the larger the database is, the better the results are [16] [15]. Thus, increasing the size of the learning database generally improves performance while increasing the diversity of the examples.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

LSSD: a Controlled Large JPEG Image Database for Deep-Learning-based Steganalysis "into the Wild"

Ruiz¹,

Yedroudj²,

Chaumont³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

For many years, the image databases used in steganalysis have been relatively small, i.e. about ten thousand images. This limits the diversity of images and thus prevents large-scale analysis of steganalysis algorithms. In this paper, we describe a large JPEG database composed of 2 million colour and grey-scale images. This database, named LSSD for Large Scale Steganalysis Database, was obtained thanks to the intensive use of "controlled" development procedures. LSSD has been made publicly available, and we aspire it could be used by the steganalysis community for large-scale experiments. We introduce the pipeline used for building various image database versions. We detail the general methodology that can be used to redevelop the entire database and increase even more the diversity. We also discuss computational cost and storage cost in order to develop images.

show abstract

Analysis of the Scalability of a Deep-Learning Network for Steganography “Into the Wild”

Cited by 11 publications

References 26 publications

How to Pretrain for Steganalysis

How to Pretrain for Steganalysis

Improved Security of a Deep Learning-Based Steganography System with Imperceptibility Preservation

LSSD: a Controlled Large JPEG Image Database for Deep-Learning-based Steganalysis "into the Wild"

Contact Info

Product

Resources

About