A basis of common approach to the development of universal steganalysis methods for digital images

Abstract

In this paper a new common approach to the organization of steganalysis in digital images is developed. New features of formal parameters defining the image are identified, theoretically grounded and practically tested. For the first time characteristics of mutual disposition of the left and right singular vectors corresponding to the largest singular value of the matrix (block of matrix) of an image and the vector composed of the singular values obtained as a result of normal singular decomposition of the matrix (block matrix) are obtained. It is shown that for the majority of the blocks of the original image (regardless of the storage format — lossy, lossless) the angle between the left (right) singular vector and the vector composed of singular numbers is determined by the angle between the n-optimal vector and the standard space basis of the corresponding dimension. It is shown that the discovered feature is violated for the mentioned formal parameters in the disturbed image. This is an indicator of integrity violation, particularly steganotransformation, and it can be used to develop new universal steganalysis methods and algorithms. Their efficiency does not depend on the specifics of steganoalgorithm used for insertion of additional information.

Keywords

• digital image
• matrix
• singular value
• singular vector
• n-optimal vector
• normal singular decomposition
• image integrity
• steganalysis method