Border-Hash Based Text Watermarking

This project implements an invisible text watermarking scheme based on a border-hash method. It allows embedding text messages into images and later extracting them.

Project Structure

• src/ - Contains the source code files
• examples/ - Contains example images for testing

How It Works

Embedding Process

1. Border Extraction & Hashing: The outermost 1-pixel-wide border of the image is used to compute a SHA-256 hash.

2. Pseudo-Random Pixel Selection: The hash is used as a seed for a pseudo-random number generator to create a reproducible sequence of interior pixel positions.

3. Message Encoding: The user's text message is converted to bits using UTF-8 encoding, and a 16-bit header containing the message length is prepended.

4. Watermark Embedding: The bits are embedded into the least significant bit (LSB) of the blue channel of the selected interior pixels.

Extraction Process

1. The same border hash is recomputed from the watermarked image.

2. The same pseudo-random sequence of pixel positions is generated.

3. The first 16 bits are extracted to determine the message length.

4. The remaining bits are extracted and converted back to text.

Installation

1. Clone this repository:

   git clone https://github.com/prtmsh/steganography-py.git
   cd steganography-py
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

Usage

Embedding a Message

python src/main.py --mode embed --input examples/input.jpg --output examples/watermarked.png --message "Your secret message"

Note: We strongly recommend using PNG format for the output file to ensure watermark integrity. The program will automatically convert the output to PNG format if another format is specified.

Extracting a Message

python src/main.py --mode extract --input examples/watermarked.png

Requirements

• Python 3
• OpenCV
• NumPy
• Pillow

Limitations

• The image must be large enough to accommodate the message.
• Maximum message size is 65,535 bytes (approx. 65KB).
• Modifications to the watermarked image (especially to the border) may prevent message recovery.
• The message is encoded in the LSB of the blue channel, which may be vulnerable to certain image manipulations.
• JPEG or other lossy compression formats will corrupt the watermark - always use PNG format.

Troubleshooting

• If extraction fails, ensure you're using the exact same watermarked image that was generated (no resizing, recompression, etc.)
• The program will display debug information about message lengths during embedding and extraction
• If the extracted text is corrupted, it means the watermark has been damaged, possibly by image compression