Simple algorithmic modifications for improving blind steganalysis performance

Schwamberger, Valentin; Franz, Matthias

doi:10.1145/1854229.1854268

Cited by 12 publications

(8 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We note that the bootstrap samples are formed "by pairs," i.e., we make sure that the pairs of cover features and the corresponding stego features are preserved. This modification, which is specific for steganalysis, is rather important as it has been shown that breaking the coverstego pairs into two sets, one of which is used for training and the other, testing, one for error estimation, may lead to a biased error estimate and, consequently, to a suboptimal performance [46], [27].…”

Section: A the Ensemblementioning

confidence: 99%

Ensemble Classifiers for Steganalysis of Digital Media

Kodovský

Fridrich

Holub

2012

IEEE Trans.Inform.Forensic Secur.

984

506

View full text Add to dashboard Cite

Abstract-Today, the most accurate steganalysis methods for digital media are built as supervised classifiers on feature vectors extracted from the media. The tool of choice for the machine learning seems to be the support vector machine (SVM). In this paper, we propose an alternative and wellknown machine learning tool -ensemble classifiers implemented as random forrests -and argue that they are ideally suited for steganalysis. Ensemble classifiers scale much more favorably w.r.t. the number of training examples and the feature dimensionality with performance comparable to the much more complex SVMs. The significantly lower training complexity opens up the possibility for the steganalyst to work with rich (high-dimensional) cover models and train on larger training sets -two key elements that appear necessary to reliably detect modern steganographic algorithms. Ensemble classification is portrayed here as a powerful developer tool that allows fast construction of steganography detectors with markedly improved detection accuracy across a wide range of embedding methods. The power of the proposed framework is demonstrated on three steganographic methods that hide messages in JPEG images.

show abstract

Section: A the Ensemblementioning

confidence: 99%

Ensemble Classifiers for Steganalysis of Digital Media

Kodovský

Fridrich

Holub

2012

IEEE Trans.Inform.Forensic Secur.

984

506

View full text Add to dashboard Cite

show abstract

“…The training set is formed by pairs of cover features and the corresponding stego features, extracted as differences with respect to their smoothed versions. This is quite important as it has been shown that breaking the cover-stego pairs may lead to a suboptimal performance [16]. Furthermore, training the classifier is a crucial issue, as the steganalyzers trained by different machine learning methods will show different performances.…”

Section: D Steganalysis Frameworkmentioning

confidence: 99%

3D mesh steganalysis using local shape features

Borş

2016

2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Steganalysis aims to identify those changes performed in a specific media with the intention to hide information. In this paper we assess the efficiency, in finding hidden information, of several local feature detectors. In the proposed 3D steganalysis approach we firstly smooth the cover object and its corresponding stego-object obtained after embedding a given watermark code. We use various operators in order to extract local features from both smoothed 3D objects. Machine learning algorithms are then used for learning to discriminate between those 3D objects which are used as carriers of hidden information and those that are not used. The proposed 3D steganalysis methodology is shown to provide superior performance to other approaches in a well known database of 3D objects.

show abstract

“…The classification was done with a 1-norm soft margin non-linear C-SVM classifier using a Gaussian kernel. The choice of the parameter C of the SVM and the width σ of the Gaussian kernel was based on the paired cross-validation procedure described in [13].…”

Section: Spam Features Type Steganalyzermentioning

confidence: 99%

Steganalysis in the Presence of Watermarked Images

Gräf

Franz

2013

2013 Ninth International Conference on Intelligent Information Hiding and Multimedia Signal Processing

Self Cite

View full text Add to dashboard Cite

Watermarked images are increasingly prevalent in the internet. Hence, any practical steganalyzer has to take the presence of watermarked images into account, particularly as potential source of false alarms due to similar embedding algorithms. In this study, we investigate the impact of watermarked images on the performance of a standard steganalyzer using two recent watermarking schemes: JPEG Compression Resistance Watermarking (JCRW) in the DCT domain [1] and Controllable Secure Watermarking (CSW) [2] in the pixel domain. Our findings show that JCRW and CSW do interfere with steganalysis. In particular, CSW-watermarked images were mostly classified as stego images thus rendering the investigated steganalyzer useless because of the excessively high false alarm rate. We propose a two-step classifier to handle this problem which achieves the same performance as the original steganalyzer, but in the presence of watermarking.

show abstract

Simple algorithmic modifications for improving blind steganalysis performance

Cited by 12 publications

References 9 publications

Ensemble Classifiers for Steganalysis of Digital Media

Ensemble Classifiers for Steganalysis of Digital Media

3D mesh steganalysis using local shape features

Steganalysis in the Presence of Watermarked Images

Contact Info

Product

Resources

About