Optimization analysis for image-based steganography using generative adversarial networks

Steganography is a method of hiding secret information within non-secret information. For the purpose of steganography, a lot of works based on convolutional neural network(CNN) were framed recent years and they showed the improvement of deep learning particularly in the field of hiding information. The major key factors that were kept in the account by those works include enhancing the capacity, invisibility, and security. In this research, a work based on steganography via generative adversarial networks was utilized to increase the invisibility and security, thus extracting that same secret image at the receiver side precisely. The focus of this research was to select the best suitable optimizer for the image based Steganography. Here, Stochastic Gradient Descent (SGD) and Adaptive Momentum (Adam) were compared and from the investigation, it was concluded that Adam optimizer performs better in handling the model to improve the hiding and revealing ability.