AI-Based and Compressed-Domain Video Steganography: A Systematic Review and Comparative Analysis (2017–2025)

Artificial intelligence has significantly transformed video steganography by improving the security, robustness, and adaptability of secret data embedding techniques. This paper presents a systematic review and comparative analysis of AI-based and compressed-domain video steganography techniques published between 2017 and 2025. Relevant studies were collected from major scientific databases and critically analyzed based on embedding domain, deep learning architecture, dataset, payload capacity, peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), mean square error (MSE), computational complexity, robustness, and resistance to steganalysis attacks. The review highlights the advantages and limitations of convolutional neural network (CNN), recurrent neural network (RNN), generative adversarial network (GAN), transformer-based, and compressed-domain approaches. Comparative analysis indicates that AI-driven techniques generally achieve superior imperceptibility and security compared to conventional methods, although higher computational requirements remain a significant challenge. The study further identifies existing research gaps, discusses practical implementation challenges, and outlines future research directions including lightweight deep learning models, explainable AI, federated learning, and real-time secure multimedia communication systems.