Steganogram removal using multidirectional diffusion in fourier domain while preserving perceptual image quality

摘要

This paper proposes a novel and efficient method for destructing any secret information hidden (steganograms) inside an image by any steganographic technique while preserving the visual quality of the images as well. The method involves an iterative multi-directional diffusion process in the Fourier domain that disrupts the stego content present in the image until the visual quality of the image does not drop below the desired threshold. Most importantly, our presented method is universal and blind and does not entail any knowledge about the employed steganography methods, the hidden message or the cover image. Simulations ran on 12600 images created by eight different state-of-the-art steganography algorithms prove that our technique succeeded in erasing from 80% to 95% of stego content on average and is superior to other similar systems. (c) 2021 Elsevier B.V. All rights reserved.