Secret sharing with multi-cover adaptive steganography

More and more studies have been dedicated to investigating secret sharing with steganography. Unfortunately, no previous work has ever reported its capability to resist steganalysis. In this paper, we pose the multi-cover adaptive steganography problem. Two secret sharing methods for natural images based on multi-cover adaptive steganography have been proposed. The secret information is adaptively shared into textured regions of covers by a spatial ±1 operation. In comparison to previous secret sharing methods, each of the proposed methods uses a simple share-constructing operation and each has lossless secret reconstruction and high quality shares. More importantly, the proposed methods are more secure in terms of resistance against state-of-the-art steganalysis techniques. In comparison to previous steganographic methods, the proposed methods hide secret bits among textured regions with different covers and are thus difficult to detect. Moreover, the proposed methods can be used to adaptively embed location-sensitive secrets (e.g., secret images) and require no stego key for extracting the encrypted message. These results also have significance in individual cover steganography. The experimental results show the effectiveness of the proposed methods.