Skip to content

Validating minimum entropy coupling steganography outputs using an image classifier network.

Notifications You must be signed in to change notification settings

blue-note/steganalysis_deep_learning

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Steganalysis of MEC Image Steganography

What is steganography?

Steganography is the practice of concealing a hidden message (a piece of media) within another piece of media. This is in contrast to cryptography, where the existence of the message is not hidden, but the content is encrypted.

In this context, a "ciphertext" is the hidden message, a "covertext" is the media that the ciphertext is hidden in, and the "stegotext" is the covertext with the ciphertext hidden inside.

Background

Earlier this year, de Witt et al showed that a minimum entropy coupling between a covertext distribution and a ciphertext distribution yields perfectly secure stegotext, ie a text with a hidden message that is undetectable by human or statistical analysis. Their experiments show that the procedure used to generate a minimum entropy coupling (iMEC) achieves minimum KL divergence as compared to other steganographic methods (A covertext and ciphertext are perfectly secure if their KL divergence is zero.)

Goal of this experiment

The goal of this experiment is to use an image classifier model trained on a stegotext dataset to validate de Witt et al's claim that stegotexts produced by iMEC are undetectable as compared to other steganographic methods such as Meteor and arithmetic coding. The steganography portion of the code is forked from their project.

About

Validating minimum entropy coupling steganography outputs using an image classifier network.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages