A Review Paper on Deep Learning-Based Image Steganography Techniques

Information Science and Engineering

The fast progress of communication and multimedia technologies means we need systems that can send data securely. Image steganography is a way to hide secret information in digital images without changing how they look. We use image steganography to conceal information within digital images while keeping the image quality good and making sure it is hard to notice. Traditional methods like Least Significant Bit (LSB) and Discrete Cosine Transform (DCT) have some problems. They cannot hide much information they are not strong and they can be easily detected [6]. Recent developments in Deep Learning, Neural Style Transfer and Generative Adversarial Networks have made image steganog-raphy systems work much better [2], [5], [7], [8]. This paper looks at all the ways that use Deep Learning for image steganography including Convolutional Neural Networks, Autoencoders, Generative Adversarial Networks, StyleGAN and Neural Style Transfer techniques. The paper also talks about ways to keep data safe with encryption like AES and ECC and ways to detect information like steganalysis methods and it looks at datasets how to measure performance, challenges and what we can do in the future. When we compare the traditional methods to the new Deep Learning methods we see that Deep Learning methods are better at keeping the image quality high and they can hide more information and they are harder to detect. Deep Learning based image steganography approaches are superior to traditional image steganography methods and they have better PSNR, SSIM, payload capacity and resistance against steganalysis.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.