Steganalysis on Internet images via domain adaptive classifier

Yang, Yong; Kong, Xiangwei; Wang, Bo; Ren, Ke; Guo, Yanqing

doi:10.1016/j.neucom.2019.04.025

Cited by 6 publications

(10 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Розміри різні, але не перевищують 500×500 пікселів (для MIRFlickr 25k 399×462 пікселі у середньому). Наприклад, у роботі [15] використовувалося 9000 випадково обраних зображень з бази MIRFlickr 25k, а в [16] -1300 випадково обраних зображень з MIRFlickr 1M.…”

Section: про вплив параметрів зображень у форматі Jpeg на точність їх стеганоаналізуunclassified

“…Так, наприклад, в роботі [17] використовувалося 49678 зображень з мережі Інтернет. В роботі [16] окрім MIRFlickr 1M використовувалися також 5000 зображень товарів з сайту https://www.amazon.cn і 2,4 мільйони зображень, завантажених з веб-сайту Flickr та за допомогою пошукової системи Google Images.…”

Section: про вплив параметрів зображень у форматі Jpeg на точність їх стеганоаналізуunclassified

See 1 more Smart Citation

About JPEG Images Parameters Impact to Steganalys Accuracy

Koshkina

2021

Cybernetics and Computer Technologies

View full text Add to dashboard Cite

Introduction. Existing examples of illegal use of computer steganography prove the need for the development of stegananalytical methods and systems as one of the most important areas of cybersecurity. The advantage of machine learning-based stegananalytical methods is their versatility: they do not rely on knowledge of the injection algorithm and can be used to detect a wide range of steganographic methods. However, before being used for detecting steganocontainers, the methods mentioned require training on containers that are determined for sure whether they contain hidden messages or not. On this stage, it is very important to understand how the parameters of containers under investigation, in particular, such a common variant as JPEG images, affect the accuracy of steganalysis. After all, the inconsistency of the source of containers is an open problem of steganalysis leading to significant decrease of accuracy of detecting hidden messages after the classifier is moved from the laboratory to the real world. The purpose of the work is investigation of influence of the content, size and quality factor of JPEG images to the accuracy of their steganalysis performed by statistical methods based on machine learning. Results. During the research the following patterns were revealed: 1) the accuracy is better when images with a close percentage of coefficients suitable for DCT concealment are used for training and control, 2) images are classified more accurately when they have a relatively small number of suitable DCT coefficients, 3) with using mixed training samples (by content or parameters) the accuracy of steganalysis deteriorates, 4) decreasing quality factor of JPEG-images leads to increasing the accuracy of their steganalysis, 5) increasing size of images increases the accuracy of their steganalysis, 6) images where desynchronization of blocks took place during preprocessing are classified more accurately, 7) the sequence of the image preprocessing operations affects the accuracy of its steganoanalysis. Conclusions. For steganography tasks – the choice of JPEG containers, taking into account revealed patterns, makes steganographic hides more resistant to passive attacks. Considering them for tasks of steganalysis allows one to interpret the obtained results more accurately. Keywords: information security, steganography, stegananalysis, intelligent computer systems, machine learning, detection accuracy.

show abstract

Section: про вплив параметрів зображень у форматі Jpeg на точність їх стеганоаналізуunclassified

About JPEG Images Parameters Impact to Steganalys Accuracy

Koshkina

2021

Cybernetics and Computer Technologies

View full text Add to dashboard Cite

show abstract

“…For the second kind of method, a domain adaptation classifier was constructed, which directly integrated domain adaptation principles as regularization terms. For instance, by adding conditional distribution into the Laplacian regularization, ARTL [26] was improved by [20]. Joint distribution adaptation and geometric structure were also integrated into the domain adaptation classifier.…”

Section: Related Workmentioning

confidence: 99%

“…First, most of the existing steganalysis schemes are based on deep learning. However, to the best of our knowledge, the cover source mismatch problems solved by [16][17][18][19][20][21] are all for handcrafted features, such as Pevny Method (PEV) [27], Rich Model [28]，CC-PEV [29], and DCTR [30]. Few works have been conducted on cover source mismatch in deep learningbased steganalysis.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Feature-guided Deep Subdomain Adaptation Network for Dataset Mismatch in Spatial Steganalysis

Zhang

Wang

et al. 2021

Preprint

View full text Add to dashboard Cite

Steganalysis aims to detect covert communication established via steganography. In recent years, numerous deep learning-based image steganalysis methods with high performance have been proposed. However, these methods tend to suffer from distinct performance degradation when cover images in the train and test set are quite different, also known as cover source mismatch. To address this limitation, in this paper, a feature-guided deep subdomain adaptation network is proposed. Initially, the predictions of the pretrained model are used as pseudo labels to divide the unlabeled samples of the target domain into different subdomains, and the distributions of the relevant subdomains are aligned by subdomain adaptation. Afterwards, since the steganalysis model may assign incorrect predictions to samples in the target domain, we integrate guiding features to make the division of subdomains more precise. The experimental results show that the proposed network is significantly better than other three networks such as Steganalysis Residual Network (SRNet), deep adaptive network (J-Net) and Deep Subdomain Adaptation Network (DSAN), when it is used to detect three spatial steganographic algorithms with a wide variety of datasets and payloads. Especially, compared with SRNet, the average accuracy of our method is increased by 5.4% at 0.4bpp and 8.5% at 0.2bpp in the case of dataset mismatch.

show abstract

Advanced Methods of Detection of the Steganography Content

Hendrych

Ličev

2019

Lecture Notes in Electrical Engineering

View full text Add to dashboard Cite

Steganografie může být využita k nelegálním aktivitám. Proto je velmi důležité být připraven. K detekci steganografického obrázku máme k dispozici techniku známou jako stegoanalýza. Existují různé typy stegoanalýzy v závislosti na tom, zda je znám originální nosič nebo zdali víme, jaký byl použit algoritmus pro vložení tajné zprávy. Z hlediska praktického použití jsou nejdůležitější metody "slepé stagoanalýzy", které zle aplikovat na obrazové soubory a jelikož nemáme originální nosič pro srovnání. Tato doktorská práce popisuje metodologii obrazové stegoanalýzy. V této práci je důležité porozumět chování cíleného steganografického algoritmu. Pak můžeme využít jeho slabiny ke zvýšení detekční schopnosti a úspěšnosti kategorizace. Primárně se zaměřujeme na prolomení steganografického algoritmu OutGuess2.0 a sekundárně na algoritmus F5. Analyzujeme schopnost detektoru, který využívá proces kalibrace, výpočtu shlukování a mělkou neuronovou síť k detekci přítomnosti steganografické zprávy na podezřelém snímku. Nový přístup a výsledky jsou sepsány v této doktorské práci. Hlavní přínosy disertační práce jsou následující: 1. Detekce stegogramů vytvořených steganografickými algoritmy OutGuess2.0 a F5. 2. Vysoká úspěšnost klasifikace -s ohledem na sensitivitu testu. 3. Invariance sensitivity testu vůči velikosti tajné zprávy. 4. Podpora různých rozlišení a barevné hloubky obrázků. 5. Aplikace mělké neuronové sítě. 6. Aplikace filtrování makrobloků JPEG obrázků pro zvýšení úspěšnosti klasifikace. 7. Publikování výzkumu -všechny prezentované přínosy jsou publikovány na mezinárodní konferenci nebo v časopise.

show abstract

Steganalysis on Internet images via domain adaptive classifier

Cited by 6 publications

References 18 publications

About JPEG Images Parameters Impact to Steganalys Accuracy

About JPEG Images Parameters Impact to Steganalys Accuracy

Feature-guided Deep Subdomain Adaptation Network for Dataset Mismatch in Spatial Steganalysis

Advanced Methods of Detection of the Steganography Content

Contact Info

Product

Resources

About