Abstract: The article consideres the accuracy of the steganoanalysis using the Weighted Stego algorithm in passive data transmission channel countermeasures using the method of embedding the spatial region of fixed digital images with the RGB color model into the smallest significant bit. The dependence of the accuracy of steganoanalysis by the Weighted Stego method on the fraction of a homogeneous background in the analyzed image was studied. The drop in the accuracy of pixel prediction in the background areas of the image is investigated using the prediction model proposed by the authors of the original Weighted Stego algorithm. The Weighted Stego steganoanalysis algorithm is investigated. The basis for the steganoanalysis algorithm is a model for predicting the pixel values of the analyzed image from adjacent pixels. To assess the effectiveness of the analysis, the BOWS2 collection was used. Embedding information is realized by changing the least significant bits of the image in the spatial domain with a payload of 3-5 %%. The effectiveness of the methods is determined taking into account the obtained truly positive, truly negative, false positive and false negative values of the classification of images. The fall in the accuracy of steganoanalysis by the Weighted Stego method is shown with an increase in the fraction of a homogeneous background in the analyzed image. The method of improvement of the pixel prediction model in the basis of Weighted Stego is proposed, which allows to level out the drop in accuracy with increasing the fraction of a homogeneous background in the analyzed image. The results of the work are useful to a specialist in the field of information protection in the tasks of detecting and countering a hidden data channel. The obtained results can be used in the development of steganoanalysis systems based on the Weighted Stego algorithm.

Abstract: The article contains the studies, which are aimed at improving the precision of steganalysis in the sphere of digital image DWT. The authors analyze the causes of inaccuracy of the modern stegoanalysis methods based upon the support vectors, then they offer the directions for improving the teaching quality. In order to improve the quality of teaching support vectors machine the authors study the interrelation between the spheres of one-dimensional and two-dimensional DWT and the influence of the changes in the coefficients of the high frequency spheres of the two-dimensional DWT upon the coefficient spheres of the one-dimensional DWT.  The steganographic influence involves the change in the value of the lower meaning bit coefficients of the DWT. Considering the study results the authors develop an original method, guaranteeing greater precision in the sphere of finding incorporated information in the high frequency areas of the two-dimensional DWT image.  In order to prove the precision of the original method, the authors compare it with some modern steganalysis methods. Experimental results of a comparative study prove that the original method provides for greater precision (generally 10-15% higher than other evaluated methods) when detecting the fact of steganographic influence in high frequency areas of HL and LH of the two-dimensional DWT. The original method also provides for the same precision in the high frequency HH area, as do other modern methods evaluated in this article.

Abstract: Countering the hidden channels of information transmission is an important task in the organization of information security. One kind of passive physical resistance methods is detection of the steganographic impact on the investigated container. The widespread use of digital still images as stegano-containers is due to their large share in total data traffic. The task of passive counteraction (steganalysis) allowing identifying the digital image with the built-in information is actually a binary classification problem. At the core of the classifier lies statistical algorithm of quantitative steganalysis for determining the amount of modified pixels in the data container. The accuracy of the algorithm directly affects the quality classification and the practical effectiveness of passive physical resistance as a whole. By effective counteraction the article refers to the ratio of probabilities between true positive classification and the probability of a false positive classification. Currently there are many statistical algorithms for quantitative steganalysis. However, there are no studies on their comparative analysis which complicates the selection of an algorithm while solving the problem of counteraction to steganography channels of information leakage. The practical effectiveness of passive physical resistance to steganography channels by inserting the least significant bits of pixel digital image also remains an open question.
The subject of the study is the effectiveness of the application of modern quantitative statistical algorithms steganalysis. Based on the results of the study the authors have formed graphics of trust regions, allowing to make a comparative assessment of the effectiveness of the passive counteraction in LSB-steganography. For the study the authors selected the following steganalysis algorithms: RS- analysis, Sample pair analysis, Difference image histogram, Triples analysis, Weighted stego-image. From the test of multiple images an image is selected. An evaluation of its capacity (defined by the maximum payload) is performed. In the experiments for this value authrs accepted the total number of pixels in the image. Steganographic effects modeled by changing the value of the least significant bit for a predetermined number of pixels (the payload). The modified image used as an input to a particular implementation of the algorithm steganalysis. The result of the algorithm is the number of changed pixels in the image. The experiments were carried out under the same conditions for all implementations of algorithms steganalysis. The main conclusions of the study is the fact that based on modern statistical steganalysis algorithms it is possible to organize an effective opposition to the passive channels with LSB steganography with embedding payload container more than 5%. Reducing the payload container of less than 5% dramatically reduces the effectiveness of the passive counteraction. A small 600x400 pixels image converted to steganography with payload of 1-2% is practically not detected by classifiers based on statistical quantitative algorithms steganalysis. Taking into account the possibility of pre-compression hidden data and matrix embedding, the considered modern algorithms for steganalysis need further improvement.