Generative Steganography

Generative steganography focuses on directly generating images or text containing hidden messages, bypassing the need for modifying existing "cover" media. Current research emphasizes developing novel architectures, including diffusion models, normalizing flows, and autoregressive models, to improve the realism of stego-media while maximizing hidden message capacity and resisting detection. This field is significant because it pushes the boundaries of data hiding techniques, impacting areas like secure communication and data protection by offering new methods for covert information transfer.