With the rapid development of computer technology, the efficient dissemination of digital images has accelerated information sharing while also intensifying the risk of copyright infringement. Traditional digital watermarking technologies struggle to meet copyright protection requirements in complex scenarios due to insufficient robustness and significant visual interference. This research proposes a technique of information hiding encryption for digital seals, which effectively embeds encrypted information into specific areas of digital seal carriers through a mask mechanism, and dynamically hides them to target images or documents, achieving a dual verification copyright protection mechanism of "seal text" and "hidden information." The experiment employed 54,673 seal images to train the steganographic network, with a test set containing 9,113 images. Results show that, hidden information extraction accuracy reached 99.81%%, maintaining stable performance even with random masking of 1%-7% of the image; residual images achieved an average peak signal-to-noise ratio (PSNR) of 30.06dB, structural similarity index (SSIM) above 0.93, and learned perceptual image patch similarity (LPIPS) below 0.01. This technology can be extended to scenarios such as image document copyright tracking systems to provide a highly robust and low-perceptibility solution for digital content copyright protection.