SPECTRAL ANALYSIS OF STEGANOGRAMS

Abstract

The data embedding algorithms in transformation domain of digital images are widely used for increasing the robustness of stegoalgorithmsto known steganalysis methods. Applying of standard transformations the digital images by steganograms forming – two-dimensional discretecosine and wavelet transforms – allows minimizing the garbling the hidden messages by image compression. We propose to use the spectral analysis methods – singular spectrum analysis and multifractal analysis – for revealing the fact of message embedding with usage of standard transform of cover image and stegodata. The advantage of such approach in comparison with statistical steganalysis methods is possibility to simultaneously investigation of several components the cover image, which allows localizing the domain of data hiding.Research of effectiveness of the spectral analysis methods was provided for the case of usage the one-stage and multi-stage embeddingmethods. It is shown that stegodata embedding leads to alteration the all components of singular value decomposition of cover images as wellas appearance the peculiar alteration of multifractal features the steganograms – increasing the cardinality of monofractal components the image. Revealed distinctive features of steganograms are stable in wide range of cover image payloads, which is simplifying of task of stegodata revealing, hidden in transformation domain of digital images.