Deep Learning-Based Video Steganography Using DCT to Improve Imperceptibility

Abstract

Steganography is an important part of modern digital contact because people need safer ways to hide data that can't be found. The Discrete Cosine Transform (DCT) is used in this study's deep learning-based video steganography design to make it harder to spot while keeping the ability to embed and being strong. The suggested method uses convolutional neural networks (CNNs) to improve the embedding process by focusing on high-frequency DCT coefficients. This is different from traditional methods, which often compromise quality or security. By putting the secret information in the luminance component (Y) of the YUV color space, the method ensures the least amount of perceptual distortion. With a Peak Signal-to-Noise Ratio (PSNR) of more than 45 dB and a Structural Similarity Index Measure (SSIM) of more than 0.96, the system works well in tests. These results show that the method has promise.