Search

This paper presents a steganalysis technique for detection of adaptive data hiding in edge areas of images with spatial least significant bit (LSB) domain systems (AE-LSB). Pixels in the edge areas which are determined by the difference values of the two consecutive pixels are embedded more secret bits than pixels located in smooth areas. The proposed steganalytic method is designed to exploit the fact that the step artifact happens in the pixel-value differencing (PVD) histograms of the stego-images due to the message embedding using AE-LSB steganography. By analyzing the step artifact in the PVD histograms of the stego-images, the statistical features that can distinguish cover-images from stego-images are obtained. Extensive experimental results indicate that our proposed scheme has superior performance compared with the prior steganalysis algorithms.

Based on the analysis of the pixel value difference (PVD) histograms of the cover and stego-images, a special steganalyzer is proposed for the edge adaptive image steganography based on least-significant-bit matching revisited (EA-LSBMR). The EA-LSBMR steganography utilizes the sharper edge regions within the cover images to embed the secret message firstly, which achieves a higher security. However, there exists unavoidable weakness: abnormal increasing at some position of the PVD histogram. The special steganslytic method is designed using the weakness of the EA-LSBMR steganography. Extensive experimental results show that the proposed method can defeat the EA-LSBMR steganography effectively.

The existing detection algorithms are difficult to obtain high detection accuracy when applied to the condition, in which the embedding rate of the stego-images is unknown. Therefore, it is very difficult to adopt the existing image steganalysis techniques in the design of a practical covert communication detection system. In this paper, combining fusion of decision-making technique", "we attempt to propose a rate-unknown image steganalysis scheme based on Dempster-Shafer(D-S) evidence theory. First, we can obtain various classifying results by establishing multi-rate training models, and then enhance these classifying results through introducing weighted coefficients according to their influences on the steganalysis. In the end, the final decision-making result is obtained by Dempster's combinational rule based on weighted coefficients. The experimental results on detecting LSB matching show that the proposed method can not only significantly improve the performance of detecting the rate-unknown stego images, but also reduce the false alarm rate. Moreover, it can control the classification effect by adjusting parameters.

In this paper, two practical forensics aided steganalyzers (FA-steganalyzer) for heterogeneous bitmap images are constructed, which can properly handle steganalysis problems for mixed image sources consisting of raw uncompressed images and JPEG decompressed images with different quality factors. The first FA-steganalyzer consists of a JPEG decompressed image identifier followed by two corresponding steganalyzers, one of which is used to deal with uncompressed images and the other is used for mixed JPEG decompressed images with different quality factors. In the second FA-steganalyzer scheme, we further estimate the quality factors for JPEG decompressed images, and then steganalyzers trained on the corresponding quality factors are used. Extensive experimental results show that the proposed two FA-steganalyzers outperform the existing steganalyzer that is trained on a mixed dataset. Additionally, in our proposed FA-steganalyzer scheme, we can select the steganalysis methods specially designed for raw uncompressed images and JPEG decompressed images respectively, which can achieve much more reliable detection accuracy than adopting the identical steganalysis method regardless of the type of cover source.