Steganography

The aim of this script is to hide a message into an image. It's a symetric encryption algorithm.

Usage

# Encode a plain text message:
$ python steganography.py -e myPicture.jpg -o encryptedPicture.bmp -m "my message"

# Encode the content of a file
$ python steganography.py -e myPicture.jpg -o encryptedPicture.bmp -f /var/myFile.txt

# Decode an image:
$ python steganography.py -d encryptedPicture.bmp -i myPicture.jpg

# Encode a plain text message with a key:
$ python steganography.py -e myPicture.jpg -o encryptedPicture.bmp -k mykey

# Decode an image with a key:
$ python steganography.py -d encryptedPicture.bmp -i myPicture.jpg -k mykey

# Generate a random image:
$ python steganography.py -g 1000x1000 -o myRandomPicture.bmp

Example

# Encode message (previously encrypted with mypasskey) into lena_encrypted.bmp based on lena.png
$ python steganography.py -e lena.jpg -o lena_encrypted.bmp -m "This is a message"

Lena.jpg (original image)

Lena.bmp (contains hidden message)

# Image is decrypted but Message is still encrypted with a passkey
$ python steganography.py -d lena.bmp -i lena.jpg
B�T�X[\�SY�UTDE���

# Passkey is needed
$ python steganography.py -d lena.bmp -i lena.jpg -k mypasskey
This is a hidden message

How does it work ?

• Encode original message into binary array (ascii) and apply it as a binary mask over the original picture
• Difference is not visible to human eye
• Addition of an additional security layer with the ability to encrypt the message before encoding it in the image (--key option). Works with a simple rotating XOR cypher algorithm. Key is previously hashed (SHA-512) to avoid hard guessing using frequency analysis
• Output image format is obviously restricted to .bmp extension. No compression allowed on the output file. However, any image format is supported as input for the original image.

Original image needs to stay private! Whereas the output image can be shown to anyone. Only someone having the original image (and the passkey if addition encryption was enabled) could decrypt the message.

Notes

Changes in bitmap image representation can be spotted by countour detection algorithms if a pixel is slighly different to the others around and does not follow any gradient rule that could naturally be explained.

As a recommendation, it's better to use a random image (full of random pixels) and not a picture taken from the web or any camera.

You can generate a random image using -g (--generate option).

Dependencies

PIL - Image