Style Transfer-based Secured Steganography: A Robust Embedding Framework

Conventional methods of information concealment accomplish this by altering carrier data, which readily leaves traces that steganalysis programs may be able to find. Both geometric and non-geometric attacks have the potential to slightly alter an image's pixel composition while it is being sent, particularly in the case of images. Provided a useful resilient image steganography method which works on style alteration to get around these problems. Our approach does not directly alter the carrier data, in contrast to conventional steganography. In this work, a dictionary mapping by establishing a correlation between image categories and binary codes has been created. It is then mapped secret data to secret images using the mapping dictionary. Later, stego images are formed by fusing the content of public photographs with the style of hidden images using techniques such as style transfer and image semantic segmentation. Information can be transmitted securely via public channels thanks to the resistance of this kind of stego picture. In order to guarantees precision, and effective retrieval of the information, The marked image is input within a reconstruction network which is already trained at the receiving end. This network is capable of successfully reconstructing the cover data and recovering the crucial information by the use of a mapping dictionary.