The Experts below are selected from a list of 888 Experts worldwide ranked by ideXlab platform
Patrick Bas - One of the best experts on this subject based on the ideXlab platform.
-
Natural Steganography in JPEG Domain With a Linear Development Pipeline
IEEE Transactions on Information Forensics and Security, 2020Co-Authors: Theo Taburet, Patrick Bas, Wadih Sawaya, Jessica FridrichAbstract:In order to achieve high practical security, Natural Steganography (NS) uses cover images captured at ISO sensitivity $ISO_{1}$ and generates Stego images mimicking ISO sensitivity $ISO_{2}>ISO_{1}$ . This is achieved by adding a Stego Signal to the cover that mimics the sensor photonic noise. This paper proposes an embedding mechanism to perform NS in the JPEG domain after linear developments by explicitly computing the correlations between DCT coefficients before quantization. In order to compute the covariance matrix of the photonic noise in the DCT domain, we first develop the matrix representation of demosaicking, luminance averaging, pixel section, and 2D-DCT. A detailed analysis of the resulting covariance matrix is done in order to explain the origins of the correlations between the coefficients of $3\times 3$ DCT blocks. An embedding scheme is then presented that takes into account all the correlations. It employs 4 sub-lattices and 64 lattices per sub-lattices. The modification probabilities of each DCT coefficient are then derived by computing conditional probabilities computed from a multivariate Gaussian distribution using the Cholesky decomposition of the covariance matrix. This derivation is also used to compute the embedding capacity of each image. Using a specific database called E1Base , we show that in the JPEG domain NS (J-Cov-NS) enables to achieve high capacity (more than 2 bits per non-zero AC DCT) and with high practical security ( $P_{\mathrm {E}}\simeq 40\%$ using DCTR and $P_{\mathrm {E}}\simeq 32\%$ using SRNet) from QF 75 to QF 100).
-
Natural Steganography in JPEG Domain with a Linear Development Pipeline
2020Co-Authors: Theo Taburet, Patrick Bas, Wadih Sawaya, Fridrich JessicaAbstract:In order to achieve high practical security, Natural Steganography (NS) uses cover images captured at ISO sensitivity $ISO_{1}$ and generates Stego images mimicking ISO sensitivity $ISO_{2}>ISO_{1}$. This is achieved by adding a Stego Signal to the cover that mimics the sensor photonic noise. This paper proposes an embedding mechanism to perform NS in the JPEG domain after linear developments by explicitly computing the correlations between DCT coefficients before quantization. In order to compute the covariance matrix of the photonic noise in the DCT domain, we first develop the matrix representation of demosaicking, luminance averaging, pixel section, and 2D-DCT. A detailed analysis of the resulting covariance matrix is done in order to explain the origins of the correlations between the coefficients of $3\times3$ DCT blocks. An embedding scheme is then presented that takes in order to take into account all the correlations. It employs 4 sub-lattices and 64 lattices per sub-lattices. The modification probabilities of each DCT coefficient are then derived by computing conditional probabilities from the multivariate Gaussian distribution using the Cholesky decomposition of the covariance matrix. This derivation is also used to compute the embedding capacity of each image. Using a specific database called E1 Base, we show that in the JPEG domain NS (J-Cov-NS) enables to achieve high capacity (more than 2 bits per non-zero AC DCT) and with high practical security ($P_{\mathrm{E}}\simeq40\%$ using DCTR from QF 75 to QF 100).Comment: 13 pages, 14 figure
-
AN EMBEDDING MECHANISM FOR NATURAL STEGANOGRAPHY AFTER DOWN-SAMPLING
2017Co-Authors: Patrick BasAbstract:Natural Steganography (NS) uses the concept of cover-source switching to provide good undetectability performances [1]. The sensor noise of the source (camera) for a given ISO sensitivity ISO1 is first modeled as an independent Gaussian distribution for each photo-site, then the embedding mimics a switch to another sensitivity ISO2(> ISO1). Because the embedding has to be performed on developed images, we investigate in this paper how to generate a Stego-Signal once the image is down-sampled. By studying different down-sampling mechanisms (sub-sampling, box down-sampling, tent down-sampling) applied on RAW images generated from a monochrome sensor, we show that the use of a convolution kernel with overlapping boundaries implies a synchronization mechanism similar to the one used by the CMD and Synch embedding schemes, but motivated here by statistical foundations. For each mechanism and scaling factor, we also compute the associated embedding rates and show that our results are in-line with experimental results previously highlighted for other steganographic schemes.
-
ICASSP - An embedding mechanism for natural steganography after down-sampling
2017 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2017Co-Authors: Patrick BasAbstract:Natural Steganography (NS) uses the concept of cover-source switching to provide good undetectability performances [1]. The sensor noise of the source (camera) for a given ISO sensitivity ISO 1 is first modeled as an independent Gaussian distribution for each photo-site, then the embedding mimics a switch to another sensitivity ISO 2 (> ISO 1 ). Because the embedding has to be performed on developed images, we investigate in this paper how to generate a Stego-Signal once the image is down-sampled. By studying different down-sampling mechanisms (sub-sampling, box down-sampling, tent down-sampling) applied on RAW images generated from a monochrome sensor, we show that the use of a convolution kernel with overlapping boundaries implies a synchronization mechanism similar to the one used by the CMD and Synch embedding schemes, but motivated here by statistical foundations. For each mechanism and scaling factor, we also compute the associated embedding rates and show that our results are in-line with experimental results previously highlighted for other steganographic schemes.
-
WIFS - Steganography via cover-source switching
2016 IEEE International Workshop on Information Forensics and Security (WIFS), 2016Co-Authors: Patrick BasAbstract:This paper proposes a new steganographic scheme relying on the principle of “cover-source switching”, the key idea being that the embedding should switch from one cover-source to another. The proposed implementation, called Natural Steganography considers the sensor noise naturally present in the raw images and uses the principle that, by the addition of a specific noise the steganographic embedding tries to mimic a change of ISO sensitivity. The embedding methodology consists in 1) designing the Stego-Signal that enable to switch from one source to another the raw domain, 2) computing the statistical distribution of the Stego-Signal in the processed domain, 3) embedding the payload in the processed domain. We show that this methodology is easily tractable whenever the processes are known and enables to embed large and undetectable payloads.
Jessica Fridrich - One of the best experts on this subject based on the ideXlab platform.
-
Natural Steganography in JPEG Domain With a Linear Development Pipeline
IEEE Transactions on Information Forensics and Security, 2020Co-Authors: Theo Taburet, Patrick Bas, Wadih Sawaya, Jessica FridrichAbstract:In order to achieve high practical security, Natural Steganography (NS) uses cover images captured at ISO sensitivity $ISO_{1}$ and generates Stego images mimicking ISO sensitivity $ISO_{2}>ISO_{1}$ . This is achieved by adding a Stego Signal to the cover that mimics the sensor photonic noise. This paper proposes an embedding mechanism to perform NS in the JPEG domain after linear developments by explicitly computing the correlations between DCT coefficients before quantization. In order to compute the covariance matrix of the photonic noise in the DCT domain, we first develop the matrix representation of demosaicking, luminance averaging, pixel section, and 2D-DCT. A detailed analysis of the resulting covariance matrix is done in order to explain the origins of the correlations between the coefficients of $3\times 3$ DCT blocks. An embedding scheme is then presented that takes into account all the correlations. It employs 4 sub-lattices and 64 lattices per sub-lattices. The modification probabilities of each DCT coefficient are then derived by computing conditional probabilities computed from a multivariate Gaussian distribution using the Cholesky decomposition of the covariance matrix. This derivation is also used to compute the embedding capacity of each image. Using a specific database called E1Base , we show that in the JPEG domain NS (J-Cov-NS) enables to achieve high capacity (more than 2 bits per non-zero AC DCT) and with high practical security ( $P_{\mathrm {E}}\simeq 40\%$ using DCTR and $P_{\mathrm {E}}\simeq 32\%$ using SRNet) from QF 75 to QF 100).
-
Natural Steganography in JPEG Domain with a Linear Development Pipeline
arXiv: Multimedia, 2020Co-Authors: Taburet Théo, Bas Patrick, Sawaya Wadih, Jessica FridrichAbstract:In order to achieve high practical security, Natural Steganography (NS) uses cover images captured at ISO sensitivity $ISO_{1}$ and generates Stego images mimicking ISO sensitivity $ISO_{2}>ISO_{1}$. This is achieved by adding a Stego Signal to the cover that mimics the sensor photonic noise. This paper proposes an embedding mechanism to perform NS in the JPEG domain after linear developments by explicitly computing the correlations between DCT coefficients before quantization. In order to compute the covariance matrix of the photonic noise in the DCT domain, we first develop the matrix representation of demosaicking, luminance averaging, pixel section, and 2D-DCT. A detailed analysis of the resulting covariance matrix is done in order to explain the origins of the correlations between the coefficients of $3\times3$ DCT blocks. An embedding scheme is then presented that takes in order to take into account all the correlations. It employs 4 sub-lattices and 64 lattices per sub-lattices. The modification probabilities of each DCT coefficient are then derived by computing conditional probabilities from the multivariate Gaussian distribution using the Cholesky decomposition of the covariance matrix. This derivation is also used to compute the embedding capacity of each image. Using a specific database called E1 Base, we show that in the JPEG domain NS (J-Cov-NS) enables to achieve high capacity (more than 2 bits per non-zero AC DCT) and with high practical security ($P_{\mathrm{E}}\simeq40\%$ using DCTR from QF 75 to QF 100).
-
deep residual network for steganalysis of digital images
IEEE Transactions on Information Forensics and Security, 2019Co-Authors: Mehdi Boroumand, Mo Chen, Jessica FridrichAbstract:Steganography detectors built as deep convolutional neural networks have firmly established themselves as superior to the previous detection paradigm – classifiers based on rich media models. Existing network architectures, however, still contain elements designed by hand, such as fixed or constrained convolutional kernels, heuristic initialization of kernels, the thresholded linear unit that mimics truncation in rich models, quantization of feature maps, and awareness of JPEG phase. In this work, we describe a deep residual architecture designed to minimize the use of heuristics and externally enforced elements that is universal in the sense that it provides state-of-the-art detection accuracy for both spatial-domain and JPEG steganography. The key part of the proposed architecture is a significantly expanded front part of the detector that “computes noise residuals” in which pooling has been disabled to prevent suppression of the Stego Signal. Extensive experiments show the superior performance of this network with a significant improvement, especially in the JPEG domain. Further performance boost is observed by supplying the selection channel as a second channel.
-
IH&MMSec - JPEG-Phase-Aware Convolutional Neural Network for Steganalysis of JPEG Images
Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security - IHMMSec '17, 2017Co-Authors: Mo Chen, Vahid Sedighi, Mehdi Boroumand, Jessica FridrichAbstract:Detection of modern JPEG steganographic algorithms has traditionally relied on features aware of the JPEG phase. In this paper, we port JPEG-phase awareness into the architecture of a convolutional neural network to boost the detection accuracy of such detectors. Another innovative concept introduced into the detector is the "catalyst kernel" that, together with traditional high-pass filters used to pre-process images allows the network to learn kernels more relevant for detection of Stego Signal introduced by JPEG steganography. Experiments with J-UNIWARD and UED-JC embedding algorithms are used to demonstrate the merit of the proposed design.
-
Steganalysis by Subtractive Pixel Adjacency Matrix
IEEE Transactions on Information Forensics and Security, 2010Co-Authors: Tomáš Pevný, Patrick Bas, Jessica FridrichAbstract:This paper presents a method for detection of steganographic methods that embed in the spatial domain by adding a low-amplitude independent Stego Signal, an example of which is least significant bit (LSB) matching. First, arguments are provided for modeling the differences between adjacent pixels using first-order and second-order Markov chains. Subsets of sample transition probability matrices are then used as features for a steganalyzer implemented by support vector machines. The major part of experiments, performed on four diverse image databases, focuses on evaluation of detection of LSB matching. The comparison to prior art reveals that the presented feature set offers superior accuracy in detecting LSB matching. Even though the feature set was developed specifically for spatial domain steganalysis, by constructing steganalyzers for ten algorithms for JPEG images, it is demonstrated that the features detect steganography in the transform domain as well.
Ali N. Akansu - One of the best experts on this subject based on the ideXlab platform.
-
Data Hiding Fundamentals and Applications - CHAPTER 3 – Communication with Side Information and Data Hiding
Data Hiding Fundamentals and Applications, 2004Co-Authors: Husrev T. Sencar, Mahalingam Ramkumar, Ali N. AkansuAbstract:The theory of data hiding has been developed mainly through employing analytical tools of communication with side information and spread spectrum communications. This is achieved by reinterpreting and adapting basic concepts such as channel, side information, and power constraints within the context of data hiding. In data hiding, the channel is the medium between the hider and the extractor, and it includes all forms of disturbances that affect the Stego Signal, which is an intelligent combination of the host Signal and the message to be conveyed. Side information available at the encoder in a communication channel model is associated with the host Signal at the embedder in the equivalent data hiding model. Power constraints in a channel communication scenario are analogous to the perceptual distortion limits that are determined based on the features of the host Signal. The bandwidth is dual to embedding Signal size, as they are both resources of the communication, and the measure of Signal-to-noise ratio corresponds to the measure of embedding-distortion to attack-distortion ratio. Correspondingly, in the dual data hiding problem, the performance of an embedding and detection technique depends on the underlying codeword generation scheme.
-
Data Hiding Fundamentals and Applications - CHAPTER 5 – Type II and Type III (Nonlinear) Data Hiding Methods
Data Hiding Fundamentals and Applications, 2004Co-Authors: Husrev T. Sencar, Mahalingam Ramkumar, Ali N. AkansuAbstract:Quantization-based data hiding methods that rely on type II and type III embedding/detection principles are studied together and compared based on three key characteristics: the type of the distortion reduction technique (postprocessing) employed in embedding, the form of demodulation used (detection function), and the optimization criterion utilized in determining the embedding/detection parameters. The data hiding rate (payload) vs robustness performance of type II methods is substantially improved by enhancing the functionality of the embedder with further processing capabilities, that is, thresholding, distortion compensation, and Gaussian mapping. In type III methods, embedding quantization is followed by a processing stage (postprocessing) that generates the Stego Signal. Performance of type III hiding methods vary based on three factors: the type of postprocessing that is incorporated with type II embedding, the choice of demodulation function used in message extraction, and the criterion used for optimizing the embedding and detection parameters. One intuitive way to evaluate the performance characteristics of type II and type III methods at varying noise levels is by considering the size of decision cells at the detector.
-
Data Hiding Fundamentals and Applications - CHAPTER 2 – Frameworks for Data Hiding
Data Hiding Fundamentals and Applications, 2004Co-Authors: Husrev T. Sencar, Mahalingam Ramkumar, Ali N. AkansuAbstract:Data hiding is a mechanism for attaching an auxiliary Signal to such information Signals. The auxiliary message Signal is inextricably tied to the information Signal and persists independently of the means by which the information Signal is disseminated. In the context of information hiding, the information Signal is called the cover Signal before the message Signal is embedded into it and the Stego Signal after the message Signal is embedded. The most important issues that arise in the study of data hiding techniques concern embedding and detecting mechanism, capacity, robustness, and transparency. The areas that contribute to the development of digital watermarking include at least the following: information and communication theory, decision and detection theory, Signal processing and transforms, and cryptography and cryptographic protocols. Each of these areas deals with a particular aspect of the data hiding problem. Information- and communication-theoretic methods deal with the data embedding side of the problem. Data hiding techniques are characterized by the embedding and extraction techniques employed. Methodologically, the proposed embedder/detector designs can be categorized into two main groups: additive spread spectrum-based methods and quantization-based methods.
-
A robust Type-III data hiding technique against cropping and resizing attacks
2002 IEEE International Symposium on Circuits and Systems. Proceedings (Cat. No.02CH37353), 2002Co-Authors: Husrev T. Sencar, Mahalingam Ramkumar, Ali N. AkansuAbstract:Proposes an oblivious information hiding method which allows watermark recovery for Signals subjected to cropping and resampling consecutively. We employ multiple embedding of a watermark Signal using a Type-III data hiding method which has a better robustness vs. rate trade off than the conventional methods under mean square error distortion measure. Cyclic autocorrelation features of the cropped-resampled Signal are used to estimate the amount of cropping. We analytically showed that within a small error range, it is possible to restore the cropped-resampled Signal to the cropped Signal. Synchronization of watermark detection in the cropped Stego-Signal is achieved by designing white noise like watermark Signals that are uncorrelated with their shifted replicas. For this purpose we use all-pass filters which are orthogonal to all their cyclic shifts. Freedom to hide data is obtained by modulating the phase of the cyclic all-pass filters. Further, we use Reed-Solomon error correcting codes both for introducing redundancy and achieving synchronization. We also address the issue of data hiding under multiple cropping attacks.
Remy Boyer - One of the best experts on this subject based on the ideXlab platform.
-
How quantization based schemes can be used in image steganographic context
Signal Processing-image Communication, 2011Co-Authors: Sofiane Braci, Claude Delpha, Remy BoyerAbstract:The quantization based embedding systems are widely used in the information hiding applications, thanks to their efficiency and simplicity. Moreover, they are known to be insecure in steganography context according to the Cachins' security definition because they distort the Stego-Signal probability density function. In this paper, we show that using the well-known spread transform (ST) combining with quantization based embedding systems provides an @e-secure Stego-system in the sense of Cachin's security definition. In other words, we show theoretically that this system preserves, in the sense of the relative entropy, the probability density function of the Stego-Signal as long as the ratio between the quantization step and the square root of the spreading factor is small. This highlights the fundamental tradeoff between these two quantities. Our theoretical conclusions are validated and illustrated on real images. Finally, a comparison with the Solanki et al. blind steganographic scheme is given.
-
MMSP - How quantization based schemes can be used in steganographic context
2009 IEEE International Workshop on Multimedia Signal Processing, 2009Co-Authors: Sofiane Braci, Claude Delpha, Remy BoyerAbstract:The quantization based embedding systems are usually used in the information hiding domain, thanks to their efficiency and simplicity. In other hand, they are known to be insecure in steganography context (according to Cachins' security definition) because they distort the Stego-Signal density function. In this paper, we show that an important kind of quantization based systems (those resulting from the combination with the spread transform and classical data hiding schemes based on quantization ) preserve the probability density function of the Stego-Signal. We prove the undetectability theoretically and show how the spread transform makes the Stego-message statistically non-dectectable in the real case.
-
MMSP - Informed Stego-systems in active warden context: Statistical undetectability and capacity
2008 IEEE 10th Workshop on Multimedia Signal Processing, 2008Co-Authors: Sofiane Braci, Claude Delpha, Remy Boyer, G. Le GuelvouitAbstract:Several authors have studied Stego-systems based on Costa scheme, but just a few ones gave both theoretical and experimental justifications of these schemes performance in an active warden context. We provide in this paper a steganographic and comparative study of three informed Stego-systems in active warden context: scalar Costa scheme, trellis-coded quantization and spread transform scalar sosta Scheme. By leading on analytical formulations and on experimental evaluations, we show the advantages and limits of each scheme in term of statistical undetectability and capacity in the case of active warden. Such as the undetectability is given by the distance between the Stego-Signal and the cover distance. It is measured by the Kullback-Leibler distance.
-
Informed Stego-systems in active warden context : Statistical undetectability and capacity
2008Co-Authors: Sofiane Braci, Claude Delpha, Remy Boyer, G. Le GuelvouitAbstract:Several authors have studied Stego-systems based on Costa scheme, but just a few ones gave both theoretical and experimental justifications of these schemes performance in an active warden context. We provide in this paper a steganographic and comparative study of three informed Stego-systems in active warden context: scalar Costa scheme, trellis-coded quantization and spread transform scalar Costa scheme. By leading on analytical formulations and on experimental evaluations, we show the advantages and limits of each scheme in term of statistical undetectability and capacity in the case of active warden. Such as the undetectability is given by the distance between the Stego-Signal and the cover distance. It is measured by the Kullback-Leibler distance.
Sofiane Braci - One of the best experts on this subject based on the ideXlab platform.
-
How quantization based schemes can be used in image steganographic context
Signal Processing-image Communication, 2011Co-Authors: Sofiane Braci, Claude Delpha, Remy BoyerAbstract:The quantization based embedding systems are widely used in the information hiding applications, thanks to their efficiency and simplicity. Moreover, they are known to be insecure in steganography context according to the Cachins' security definition because they distort the Stego-Signal probability density function. In this paper, we show that using the well-known spread transform (ST) combining with quantization based embedding systems provides an @e-secure Stego-system in the sense of Cachin's security definition. In other words, we show theoretically that this system preserves, in the sense of the relative entropy, the probability density function of the Stego-Signal as long as the ratio between the quantization step and the square root of the spreading factor is small. This highlights the fundamental tradeoff between these two quantities. Our theoretical conclusions are validated and illustrated on real images. Finally, a comparison with the Solanki et al. blind steganographic scheme is given.
-
MMSP - How quantization based schemes can be used in steganographic context
2009 IEEE International Workshop on Multimedia Signal Processing, 2009Co-Authors: Sofiane Braci, Claude Delpha, Remy BoyerAbstract:The quantization based embedding systems are usually used in the information hiding domain, thanks to their efficiency and simplicity. In other hand, they are known to be insecure in steganography context (according to Cachins' security definition) because they distort the Stego-Signal density function. In this paper, we show that an important kind of quantization based systems (those resulting from the combination with the spread transform and classical data hiding schemes based on quantization ) preserve the probability density function of the Stego-Signal. We prove the undetectability theoretically and show how the spread transform makes the Stego-message statistically non-dectectable in the real case.
-
MMSP - Informed Stego-systems in active warden context: Statistical undetectability and capacity
2008 IEEE 10th Workshop on Multimedia Signal Processing, 2008Co-Authors: Sofiane Braci, Claude Delpha, Remy Boyer, G. Le GuelvouitAbstract:Several authors have studied Stego-systems based on Costa scheme, but just a few ones gave both theoretical and experimental justifications of these schemes performance in an active warden context. We provide in this paper a steganographic and comparative study of three informed Stego-systems in active warden context: scalar Costa scheme, trellis-coded quantization and spread transform scalar sosta Scheme. By leading on analytical formulations and on experimental evaluations, we show the advantages and limits of each scheme in term of statistical undetectability and capacity in the case of active warden. Such as the undetectability is given by the distance between the Stego-Signal and the cover distance. It is measured by the Kullback-Leibler distance.
-
Informed Stego-systems in active warden context : Statistical undetectability and capacity
2008Co-Authors: Sofiane Braci, Claude Delpha, Remy Boyer, G. Le GuelvouitAbstract:Several authors have studied Stego-systems based on Costa scheme, but just a few ones gave both theoretical and experimental justifications of these schemes performance in an active warden context. We provide in this paper a steganographic and comparative study of three informed Stego-systems in active warden context: scalar Costa scheme, trellis-coded quantization and spread transform scalar Costa scheme. By leading on analytical formulations and on experimental evaluations, we show the advantages and limits of each scheme in term of statistical undetectability and capacity in the case of active warden. Such as the undetectability is given by the distance between the Stego-Signal and the cover distance. It is measured by the Kullback-Leibler distance.