The goal of this tool is to provide watermark services to owners of copyrighted artwork to prevent tampering and edition without detection. The tool started as a project of the Forensics Cyber-Security course at Instituto Superior Técnico, and now we are open source it.
- Major code refactoring to add a more permanent structure to the code.
- Implement test-first development.
- Write complete documentation.
- Tweak the implemented algorithms to erase some of the restrictions that we currently have.
- Develop an user interface.
- Get Python 2.7 with pip.
- Install virtualenv
pip install virtualenv - Run
virtualenv venv - Run
source venv/bin/activateto activate virtual environment - Run
pip install -r requirements.txtto install project dependencies
Make sure you have the virtual environment activated before issuing any command, or ImportErrors will occur.
The template to perform an operation is the following:
python run.py <operation> -a <algorithm> <image_file> <watermark>
where the <watermark> parameter is optional depending on the algorithm and operation.
The operations available are:
-
<operation> = embed: Embed a watermarkpython run.py embed -a cox img/lena.pngpython run.py embed -a dwt img/lena.png img/milk.pngpython run.py embed -a recover img/lena.png -
<operation> = extract: Extract a watermarkpython run.py extract -a cox wm-img/lena.png wm-img/lena_wm.jsonwhere
wm-img/lena.pngis the watermarked image andwm-img/lena_wm.jsonis the watermark file produced by embed operation.python run.py extract -a dwt wm-img/lena.pngwhere
wm-img/lena.pngis the watermarked imagepython run.py extract -a recover wm-img/lena.png wm-img/lena_kwhere
wm-img/lena.pngis the watermarked image andwm-img/lena_kis a file that contains the key to build the sequence of blocks used in the embed phase of the recovery algorithm. -
<operation> = benchmarks: Run the benchmarks
See next section.
The benchmarks automate the attacks on watermarked images. We currently implement these attacks:
- Add 20% of contrast
- Unsharp mask (a technique to increase sharpness in images).
- Mode Filter (a type of low-pass filter).
- Median Filter (a type of low-pass filter).
- Rotate 3 degrees counterclockwise.
- Add noise.
- Add gausian blur.
- Use JPEG compression with 10% quality.
- Draw white squares at the middle of the image (sizes: 5%, 15% and 30%).
- Double the size of the image, then reduce to half again.
- Reduce the size of the image to half, then double the size.
To run the benchmarks, issue the following command:
python run.py benchmarks -a <algorithm> <original_image> <watermark>
The <algorithm> parameter refers to one of the algorithms that is implemented, the <original_image> parameter
to the path of the original image. The <watermark> is only used in DWT algorithm in the current version and is
the path of the watermark image.
We currently do not support automatic benchmarks for the recover algorithm.
Examples:
python run.py benchmarks -a cox img/lena.png
python run.py benchmarks -a dwt img/lena.png img/milk.png