Chunhua Chen

A Markov process based approach to effective attacking JPEG steganography

In this paper, a novel steganalysis scheme is presented to effectively detect the advanced JPEG steganography. For this purpose, we first choose to work on JPEG 2-D arrays formed from the magnitudes of quantized block DCT coefficients. Difference JPEG 2-D arrays along horizontal, vertical, and diagonal directions are then used to enhance changes caused by JPEG steganography. Markov process is applied to modeling these difference JPEG 2-D arrays so as to utilize the second order statistics for steganalysis. In addition to the utilization of difference JPEG 2-D arrays, a thresholding technique is developed to greatly reduce the dimensionality of transition probability matrices, i.e., the dimensionality of feature vectors, thus making the computational complexity of the proposed scheme manageable. The experimental works are presented to demonstrate that the proposed scheme has outperformed the existing steganalyzers in attacking OutGuess, F5, and MB1.

Steganalysis based on multiple features formed by statistical moments of wavelet characteristic functions

In this paper, a steganalysis scheme based on multiple features formed by statistical moments of wavelet characteristic functions is proposed. Our theoretical analysis has pointed out that the defined n-th statistical moment of a wavelet characteristic function is related to the n-th derivative of the corresponding wavelet histogram, and hence is sensitive to data embedding. The selection of the first three moments of the characteristic functions of wavelet subbands of the three-level Haar wavelet decomposition as well as the test image has resulted in total 39 features for steganalysis. The effectiveness of the proposed system has been demonstrated by extensive experimental investigation. The detection rate for Cox et al.s non-blind spread spectrum (SS) data hiding method, Piva et al.s blind SS method, Huang and Shis 8×8 block SS method, a generic LSB method (as embedding capacity being 0.3 bpp), and a generic QIM method (as embedding capacity being 0.1 bpp) are all above 90% over all of the 1096 images in the CorelDraw image database using the Bayes classifier. Furthermore, when these five typical data hiding methods are jointly considered for steganalysis, i.e., when the proposed steganalysis scheme is first trained sequentially for each of these five methods, and is then tested blindly for stego-images generated by all of these methods, the success classification rate is 86%, thus pointing out a new promising approach to general blind steganalysis. The detection results of steganalysis on Jsteg, Outguess and F5 have further demonstrated the effectiveness of the proposed steganalysis scheme.

A natural image model approach to splicing detection

Image splicing detection is of fundamental importance in digital forensics and therefore has attracted increasing attention recently. In this paper, we propose a blind, passive, yet effective splicing detection approach based on a natural image model. This natural image model consists of statistical features extracted from the given test image as well as 2-D arrays generated by applying to the test images multi-size block discrete cosine transform (MBDCT). The statistical features include moments of characteristic functions of wavelet subbands and Markov transition probabilities of difference 2-D arrays. To evaluate the performance of our proposed model, we further present a concrete implementation of this model that has been designed for and applied to the Columbia Image Splicing Detection Evaluation Dataset. Our experimental works have demonstrated that this new splicing detection scheme outperforms the state of the art by a significant margin when applied to the above-mentioned dataset, indicating that the proposed approach possesses promising capability in splicing detection.

Image steganalysis based on moments of characteristic functions using wavelet decomposition, prediction-error image, and neural network

In this paper, a general blind image steganalysis system is proposed, in which the statistical moments of characteristic functions of the prediction-error image, the test image, and their wavelet subbands are selected as features. Artificial neural network is utilized as the classifier. The performance of the proposed steganalysis system is significantly superior to the prior arts.

A machine learning based scheme for double JPEG compression detection

Double JPEG compression detection is of significance in digital forensics. We propose an effective machine learning based scheme to distinguish between double and single JPEG compressed images. Firstly, difference JPEG 2D arrays, i.e., the difference between the magnitude of JPEG coefficient 2D array of a given JPEG image and its shifted versions along various directions, are used to enhance double JPEG compression artifacts. Markov random process is then applied to modeling difference 2-D arrays so as to utilize the second-order statistics. In addition, a thresholding technique is used to reduce the size of the transition probability matrices, which characterize the Markov random processes. All elements of these matrices are collected as features for double JPEG compression detection. The support vector machine is employed as the classifier. Experiments have demonstrated that our proposed scheme has outperformed the prior arts.

Effective steganalysis based on statistical moments of wavelet characteristic function

In this paper, an effective steganalysis based on statistical moments of wavelet characteristic function is proposed. It decomposes the test image using two-level Haar wavelet transform into nine subbands (here the image itself is considered as the LL/sub 0/ subband). For each subband, the characteristic function is calculated. The first and second statistical moments of the characteristic functions from all the subbands are selected to form an 18-dimensional feature vector for steganalysis. The Bayes classifier is utilized in classification. All of the 1096 images from the CorelDraw image database are used in our extensive experimental work. With randomly selected 100 images for training and the remaining 996 images for testing, the proposed steganalysis system can steadily achieve a correct classification rate of 79% for the non-blind Spread Spectrum watermarking algorithm proposed by Cox et ai, 88% for the blind Spread Spectrum watermarking algorithm proposed by Piva et ai, and 91% for a generic LSB embedding method, thus indicating significant advancement in steganalysis.

Statistical Moments Based Universal Steganalysis using JPEG 2-D Array and 2-D Characteristic Function

Owing to the popular usage of JPEG images, the steganographic tools for JPEG images emerge increasingly nowadays, among which OutGuess, F5, and the model based steganography are the most advanced. Advancing the previous work, we present a new universal steganalysis method based on statistical moments derived from both image 2-D array and JPEG 2-D array in this paper. In addition to the first order histogram, the second order histogram is considered. Consequently, the moments of 2-D characteristic functions are also used for steganalysis. Extensive experimental works have shown that the proposed method outperforms in general the prior-arts of steganalysis methods in attacking the three aforesaid steganographic schemes.

Steganalysis Versus Splicing Detection

Aiming at detecting secret information hidden in a given image using steganographic tools, steganalysis has been of interest for years. In particular, universal steganalysis, not limited to attacking a specific steganographic tool, is of extensive interests due to its practicality. Recently, splicing detection, another important area in digital forensics has attracted increasing attention. Is there any relationship between steganalysis and splicing detection? Is it possible to apply universal steganalysis methodologies to splicing detection? In this paper, we address these intact and yet interesting questions. Our analysis and experiments have demonstrated that, on the one hand, steganography and splicing have different goals and strategies, hence, generally causing different statistical artifacts on images. However, on the other hand, both of them make the touched (stego or spliced) image different from the corresponding original (natural) image. Therefore, natural image model based on a set of carefully selected statistical features under the machine learning framework can be used for steganalysis and splicing detection. It is shown in this paper that some successful universal steganalytic schemes can make promising progress in splicing detection if applied properly. A more advanced natural image model developed from these state-of-the-art steganalysis methods is thereafter presented. Furthermore, a concrete implementation of the proposed model is applied to the Columbia Image Splicing Detection Evaluation Dataset, which has achieved an accuracy of 92%, indicating a significant advancement in splicing detection.

Steganalyzing Texture Images

A texture image is of noisy nature in its spatial representation. As a result, the data hidden in texture images, in particular in raw texture images, are hard to detect with current steganalytic methods. We propose an effective universal steganalyzer in this paper, which combines features, i.e., statistical moments of 1-D and 2-D characteristic functions extracted from the spatial representation and the block discrete cosine transform (BDCT) representations (with a set of different block sizes) of a given test image. This novel scheme can greatly improve the capability of attacking steganographic methods applied to texture images. In addition, it is shown that this scheme can be used as an effective universal steganalyzer for both texture and non-texture images.

Effect of Recompression on Attacking JPEG Steganographic Schemes An Experimental Study

In the implementation of a few JPEG steganographic schemes such as OutGuess and F5, an additional JPEG compression may take place before data embedding. The effect of this recompression on the performances of steganalyzers is experimentally studied and reported in this paper. Through a group of carefully designed experimental works, we show that the training and testing procedures adopted in classification are of great importance. An improper training and testing procedure may lead to poor steganalysis performance even for a powerful steganalyzer or an accurate performance comparison. Some other informative observations are presented in the paper as well.