Enhancing Image Steganography with Deep Learning and Cryptographic QR Code Embedding

Steganography in images refers to the art and science of concealing data within digital images. This technique enables embedding messages or sensitive information into an image file in such a way that their presence is imperceptible to an ordinary observer. Various steganography methods aim to securely embed the maximum amount of confidential information within the carrier image, while ensuring that the quality of the steganographic image does not significantly degrade and remains imperceptible to the human visual system. This study proposes a high-capacity image steganography method leveraging deep learning techniques. The model employs multiple encoding mechanisms to encode text data, which is subsequently encrypted into a QR code to enhance security. A convolutional neural network (CNN) is utilized to determine optimal parameter ratios for embedding the QR code into the carrier image. The steganography process is further refined using discrete cosine transform (DCT). Additionally, a wavelet compression technique is applied to reduce the volume of the stego image while preserving the embedded data. This method enhances the image's resistance to tracking attacks while preventing any damage to the confidential information hidden within the carrier image. Finally, the Group Method of Data Handling (GMDH) neural network, combined with evaluation metrics such as MSE, RMSE, PSNR, SSIM, and payload capacity, is employed to assess the accuracy of the proposed method during stego image formation. The experimental results indicate that the proposed method achieves a high accuracy rate of 99.69% in embedding barcode encryption within images.