This research proposes a cover-reproducible steganography method using deep generative models. It explores:

• New techniques for embedding and extracting messages in covert signals.
• Use of deep learning for generating reproducible cover signals for secure communication.

The ability to securely and covertly communicate using generative models could expand the boundaries of secure communications and privacy protection.