Color Image Steganalysis Based On Steerable Gaussian Filters Bank

Abstract

This article deals with color images steganalysis based on machine learning. The proposed approach enriches the features from the Color Rich Model by adding new features obtained by applying steerable Gaussian filters and then computing the co-occurrence of pixel pairs. Adding these new features to those obtained from Color-Rich Models allows us to increase the detectability of hidden messages in color images. The Gaussian filters are angled in different directions to precisely compute the tangent of the gradient vector. Then, the gradient magnitude and the derivative of this tangent direction are estimated. This refined method of estimation enables us to unearth the minor changes that have occurred in the image when a message is embedded. The efficiency of the proposed framework is demonstrated on three stenographic algorithms designed to hide messages in images: S-UNIWARD, WOW, and Synch-HILL. Each algorithm is tested using different payload sizes. The proposed approach is compared to three color image steganalysis methods based on computation features and Ensemble Classifier classification: the Spatial Color Rich Model, the CFA-aware Rich Model and the RGB Geometric Color Rich Model.

An Empirical Study of Steganography and Steganalysis of Color Images in the JPEG Domain

“…We can set the message size P and choose β, and α and γ will be calculated using (2) and (1) respectively. In a more conventional way, we can set the embedding rate α and choose β, and P and γ will be computed using (2) and (1) respectively. If we worked on grayscale images, we would set β = 0 which means P = αN Y .…”

“…More accurately, whenever embedding is realized on color numerical images in the pixel domain (usually for steganalysis purposes), it is most of the time implemented independently on each component [1,4,5] but more advanced schemes use a synchronization strategy to have more coherent embedding changes across the different channels [13]. Popular implementations of color JPEG steganography such as F5 [15] or J-UNIWARD [8] alter DCT coefficients without taking into account their related color channels or the related component statistics.…”

“…Regarding spatial steganography and RGB components, Goljan et al developed the Spatial and Color Rich Models (SCRMQ1) [5], which can be seen as a spatial extension of the Spatial Rich Model (SRM) [3] that uses Color Co-occurrence Matrices. These features, as well as other specific feature sets, such as the Gaussian filter bank features [1] (which is an extension of the SCRMQ1), or the CFA-aware features [4], have been particularly designed for color pixel-based steganography. Recently a steganalysis scheme using deep convolutional networks [16] has been proposed, it uses 3 disjoint layers (one for each color components) that are pooled together in the next layers.…”

Rethinking the high capacity 3D steganography: Increasing its resistance to steganalysis

“…However, the steganography's resistance to the 3D steganalysis was not considered in [9]. The existing image steganalytic methodology [14,15] cannot be applied to 3D meshes, because unlike 3D objects, images are represented on regular lattices. The first 3D steganalytic algorithm was proposed in [10].…”

3D Steganalysis Using Laplacian Smoothing at Various Levels

“…In order to be used by many applications, 3D objects are increasingly transferred shared between users through clouds or mobile media. When compared to the steganalysis research on other media, such as images [15,20,1,30,17], video [24,29] and audio signals [18,19], the steganalysis for 3D objects is much less developed, resulting in a lower likelihood of identifying the information hidden in 3D objects. Many 3D information hiding algorithms have already been proposed [5,3,28,16,2,8,12].…”