An information-theoretic approach to steganography and watermarking

Abstract

A theoretical model for steganography and digital watermarking is presented, which includes a stego encoder, a stego channel and a stego decoder. The first part states the basic steganographic and watermarking problems in terms of mutual information of the involved quantities such as the secret message, the stego message and the modified stego message. General lower bounds on the robustness-related mutual information are derived. In the second part, perfect steganography is considered and some new schemes are presented that achieve perfect secrecy and provide robustness against some attacks. In the last part, the robustness of some simplistic schemes is evaluated by tight lower bounds on the robustness-related mutual information. From these bounds, two criteria for robust embedding are derived.