Secure steganography using randomized cropping

In this paper, a novel steganographic scheme is proposed, where embedding is done adaptively in image regions with high level of high frequency component. The steganalytic detection performance becomes poorer as these high frequency components mask the steganographic embedding noise. The security of the proposed scheme is further increased by separating the embedding domain from the steganalytic domain. This separation is done by randomizing the embedding domain using a new concept called randomized cropping. State of the art spatial domain steganalyzers are considered to evaluate the security of the proposed scheme. The LSB matching algorithm is used for steganographic embedding. It is shown experimentally that the LSB matching algorithm wrapped with proposed scheme performs comparatively better than simple LSB matching scheme, against the steganalytic attacks under consideration.