Steganography is a cryptographic process where we attempt to hide a file or message inside of another vessel file, usually an image or audio file, such that the vessel file appears unchanged. A traditional steganographical algorithm, such as Least Significant Bit steganography does this by encoding each bit for a file inside the least significant bit of the vessel file.
Our project attempts to use neural networks to learn their own novel form of steganography.
- Write a simple function to return an image capacity based on its dimensions
- Pad file batches with dummy bits so that each image batch has an associated message batch
- Pad image with whitespace so that trim is included in batches to encode. Trim should not count towards image capacity as padded whitespace will be removed from embedded image.
- Bring back Eve as an adversary attempting to recover bits, Alice and Bob's goal should be a 50% bit reconstruction rate while Eve hopes to achieve 100%
Visual Studio Code
Click the Extensions panel Search for "@recommended" Install the Workspace Recommended extensions
pip install -r requirements.txt
Click Tasks -> Run Tasks -> Install requirements