shujaatbakhsh25 / bestgroupever

FE 595 project repo

Digital Watermarking of Images

BestGroupEver - Final Project | Financial Technology (FE595)

Preety Vandana | William Long | Shujaat Bakhsh | Gaurav Nagal

Overview

This is the code for digital watermarking of images. We have built a model which embeds watermark to images and decodes it, which can be used to trace the original user with whom it was shared. This technology helps in protecting the author's work and rights, and ultimately preventing piracy.
This project's crux lies in the idea of a Canary Trap. A set of images is watermarked with a code unqiue to each user and then the media is distributed. In case there is a leak of this media, the embedded watermark can be decoded from the leaked image and connected to the person who leaked it.

Requirements

pip install -r requirements.txt

Usage

• Once dependencies are installed, download the project directory.
• Make sure the current directory is the project root directory.
• Put the images to be watermarked in src folder.
• Run python3 main.py (main.py program calls Add_WatermarkV3.py to generate a list of unique images/arrays for the users defined in users.csv).
• That's it! Program loads all the orginal images from src folder and stores the watermarked images to dest folder consisting of individual users.
• We can identify the original user with which the image file was shared by reading and decoding the watermarked image.
• The program prints out results of the decoding process to stdout.

Results

Here are some of the results comparing original and watermarked images.
The left one is original image and right one is watermarked image.

bestgroupever's Issues

Read watermark

Here's what we need to for the read_watermark.py file:

• take in an watermarked image and the og image.
• return the difference between the two as a new image (this should be the watermark on a blank background.)
• read that image and decode the watermark.
• return the code in that watermark.

Please update reedsolo

Hello,

I am the current maintainer of reedsolo.

We have recently pushed major updates which increase speed and stability, but the API had to be changed.

If you are still using reedsolo, please ensure that you handle the 3 variables returned by reedsolo.

Before:

>>> rsc = RSCodec(12)
>>> tampered_msg = b'heXlo worXd\xed%T\xc4\xfdX\x89\xf3\xa8\xaa'
>>> decoded_msg = rsc.decode(tampered_msg)

Now:

>>> rsc = RSCodec(10)
>>> tampered_msg = b'heXlo worXd\xed%T\xc4\xfdX\x89\xf3\xa8\xaa'
>>> rmes, rmesecc, errata_pos = rsc.decode(tampered_msg)

Things we need to do for Main

Here's what we need to do for Main.py:

• Be able to take in an image and a list of users.

• be able to make a unique watermark for each user.

• be able to save all those watermarked images.

• be able to take in an image and return the user it was sent to.