Polar Codes for Multilevel Steganography

In this paper, we use polar codes to design the multilevel steganography for both the blind watermarking and non-blind watermarking problems, respectively. The proposed steganography divides the secret information into two parts: the major part that carries the most important message and the minor part that is supplementary. When the cover sequence embedded with the secret information is slightly damaged, both parts can be extracted successfully; when the damage gets more severe, only the major part will be extracted. By systematically combining channel polarization and source polarization techniques, we show that our schemes can approach the embedding capacity from both theoretical and numerical simulation perspectives.