Chapter 12 – Steganography

Publisher Summary This chapter focuses on a specific data-hiding application—steganography. As opposed to digital watermarking, the main property of steganography is statistical undetectability of embedded data. The payload is usually unrelated to the cover Work, which only serves as a decoy. The information-theoretic definition of steganographic security (Cachin's definition) is the most widely used definition in practice. It is usually applied in a simplified form by accepting a model for the cover Work. Secure steganographic schemes must take into account steganalytic methods. One possibility is to replace the embedding operation of LSB flipping (F5) to avoid introducing easily detectable artifacts. Another possibility is to mask the embedding distortion as a naturally occurring phenomenon, such as during image acquisition (stochastic modulation). Alternatively, one can design schemes that preserve some vital statistical characteristics of the cover image (OutGuess) or a model of the cover that is recoverable from the stego Work (model-based steganography). In a typical steganographic scheme, the placement of embedding changes (the selection rule) is shared between the sender and the recipient. However, there are many situations when this information cannot be shared, such as in adaptive steganography, selection rules determined from side information, or in public-key steganography. The problem of nonshared selection rules is equivalent to writing in memory with defective cells and can be efficiently approached using sparse linear codes, known as LT codes.