Efficient cover image selection based on spatial block analysis and DCT embedding

In steganography, the cover medium is widely treated as a mere container for the embedded information, even though it affects the stego-image quality, security, and robustness. In addition, there is no consensus on the characteristics of a suitable cover image. In this work, we introduce and practically prove the most suitable cover image (MSCI) framework to automatically select a cover image for a given secret image. This paper proposes choosing the most suitable cover from a set of images based on two steps. First, a set of cover images is filtered based on relative entropy and a histogram in order to identify the most suitable candidates. Second, the local block pixel intensity features of the candidates are analyzed to select the most suitable cover image. Furthermore, cover image local blocks were optimized, using rotation and flipping, during the embedding process to further improve stego-image representation. The proposed framework demonstrated high visual image quality when compared with existing solutions. Steganalysis tests indicated that the proposed solution for cover selection provided an increased resistance to modern steganalyzers with up to 30% lowered detection rate, which improved security.