Optimized Reversible Data Hiding with CNN Prediction and Enhanced Payload Capacity

The security of digital data is guaranteed while processing and storage in encrypted cloud storage thanks to Reversible Data Hiding (RDH). During this process, users stop cloud providers (CP) from accessing image data. This paper presents a high-capacity RDH technique for safe media transmission. Based on pixel parity, the image is first divided into two sub-images. First sub-image is made up of odd pixels, and second is made up of even pixels. The Most Significant Bit (MSB) of the pixels for embedding is predicted using a Convolutional Neural Network (CNN) predictor. The secret information is encoded in the embedding using a plus/minus one technique. The cover image can be recovered with the help of an encryption key that is shared by the sender and recipient. CNN predictors use the plus-minus one technique to optimise the embedding strategy. They are renowned for their improved receptive mechanisms and increased utilisation of neighbouring pixels. Promising results are obtained by combining the CNN predictor with the plus-minus one embedding strategy, which outperforms the existing methods in terms of both visual quality and embedding capacity. For example, the average Peak Signal-to-Noise Ratio (PSNR) value for 15,000 bits on the Kodak dataset reaches an astounding 67 dB.