Tensorflow Implementation of Hiding Images in Plain Sight: Deep Steganography (unofficial)
Steganography is the science of Hiding a message in another message. In this case, a Picture is hidden inside another picture using Deep Learning.
Blog Post on it can be found here
The dependencies can be installed by using
pip install -r requirements.txt
This will install the tensorflow CPU version by default. If you would like to use your GPU , you can do
pip install --force-reinstall tensorflow-gpu
This basically reinstalls the gpu version of tensorflow for your system.
The Framework takes in Two images. One is the secret image(extreme right) and another is the cover image(extreme left).
The goal is to 'hide' the secret image in the cover image such that only the cover image is visible. This is the covered image(Center Left)
Then , this hidden image can be passed to a Revealing network, which can get the hidden image back(Center Right).
Prep Net, Hide Net , and Reveal net have the same convolutional block structure. Therefore, in the image , only the reveal network is shown, and prep/hide networks are collapsed (to make the image fit).
In order to test the working of the networks , we have written a browser inplementation using Keras JS. You can find it here:
https://harveyslash.github.io/Deep-Steg-JS/ Due to the extremely high computational resources that it requires, it takes about 3 minutes to setup.