Relabel, Scramble, Synthesize:A Novel Coverless Steganography Approach via Collage Image

In this age of digital information, most correspon-dences are now transmitted through online channels, including private, confidential, and secret messages. Hence, a significant concern in such correspondence is the threat of eavesdropping. This work proposes a novel framework of coverless steganog-raphy by secret representations relabelling and positions scram-bling to increase the complexity of a secret message encoding. This is followed by an ACGAN-based model implementation, namely StArtGAN, that is used in tandem with duotone effects to construct collage images from the encoded message, so as to produce a medium of inconspicuous communication. The decoding process then leverages on the discriminator of the StArtGAN model as the unique decoder to ensure the integrity and accurate recovery of the secret message. Investigations in this paper show that the proposed framework is scalable and holds potential to reduce the risk of eavesdropping in practical applications.