A high-capacity Coverless Image Steganography based on the Local Binary Pattern feature and WGAN model

Abstract

By concealing a hidden message inside a cover image without changing it, the idea of coverless image steganography increases the image's resistance to image steganalysis tools. This strategy outperforms conventional high-capacity techniques that use several cover images as stego images, making them unsuitable for real-time applications. The study explores the effectiveness of coverless information concealing by using Wasserstein Generative adversarial networks and the Local Binary Pattern (LBP) feature to transmit sensitive data using a single cover image.In order to generate a sufficient number of distinct hash codes, the embedding process requires overlapping image blocks. To do this, an image must be divided into blocks, the hash code must be determined, and the secret message must be linked to image blocks that have the same hash code. To speed up the embedding process, a lookup table is created. A meaningful, uncorrelated image is created by feeding the generated stego image into the Generative Adversarial Networks model. This image is then transmitted to the recipient, where it is processed by the Wasserstein Generative Adversarial Networks model to reconstruct the stego image. Research indicates that in order to produce a sufficient number of unique hash codes, overlapping image blocks are required.A single image would not be sufficient to finish the embedding process without this overlap. Sending a meaning-normal image that is unrelated to the stego image can accomplish the same goal as sending the stego image. Compared to existing covert information hiding strategies, this suggested method offers a larger capacity for information concealing and is more resilient to attacks during image processing. Additionally, this technique works quickly and provides better defense against detection tools.