An Adaptive Reversible Data Hiding Scheme using Image Processing and Deep Learning Techniques

Image steganography is the process of hiding data from digital photos without changing how they look on the outside. It functions as a covert mode of communication, giving access to confidential data without raising the suspicions of opponents or unauthorized parties. It provides an extra degree of secrecy and security for digital communications. Prediction-error expansion (PEE) is an extensively used method in the type of reversible data hiding (RDH) that hiding data with intact image quality in the prediction errors. High-accuracy prediction algorithms are typically essential for a PEE-based RDH implementation to be successful. Current PEE-based Reversible Data Hiding (RDH) systems frequently use a single forecasting approach, typically with worldwide optimization, but pay little attention to the local neighborhoods' pixel distribution features. This study aims to analyze the most recent RDH techniques and investigate image processing and deep learning (DL) techniques for reversible data hiding (DH) schemes. RDH is a difficult topic in general which has potential uses in today's digital world. Reversible data hiding techniques properly restore the cover medium while safely transferring secret data (SD). The goal of RDH techniques is to achieve both large capacity and tunable quality. Even though reversible data hiding has been the subject of indepth research recently, there hasn't yet been a thorough literature analysis to identify research gaps and potential future approaches.