/Vector-based-RDHEI-via-Multi-MSB-Replacement

Primary LanguagePythonMIT LicenseMIT

Vector-based RDHEI via Multi-MSB Replacement

This is the source code for IEEE TCSVT journal paper Vector-based Efficient Data Hiding in Encrypted Images via Multi-MSB Replacement.

Getting Started

These instructions below will guide you on running this project on your local machine for development and testing purposes.

Note For addtional information and references, please check the Appendix for this paper.

Warning Both of the EMR and LMR methods can run perfectly in UNIX-based systems (such as MacOS) since we pre-compiled the JBIG-KIT compression source code to executable files. It will have some issues when running the LMR method in the Windows due to the source .c files are not compiled to .exe in the provided code. In short, if the JBIG-KIT software can be compiled and run in your local machine, then the LMR should run seamlessly as well. Moreover, the EMR-RDHEI method is not depending on any compression kit, thus it should run successfully without any issue.

File structure

There is the file structure provided below:

RDHEI-EMR-LMR
	└── source_code
			└── assets // for all testing images
			└── outputs // for all the testing outputs
			└── temp // store temporary files, not important
			└── tools // JBIG-KIT executable software
			└── utils
			└── auto.py // for automated testing of 10,000 images in BOWS2 database
			└── demo.py // for demo a specific image purpose
			└── eval.py // for security and image resolution evaluation purpose
			└── entity.py
			└── EMR.py
			└── LMR.py

	└── environemnt.yml // use conda/miniconda to recreate the environment
	└── LICENSE
	└── README.md
	└── requirements.txt // use pip to recreate the enviroment

Prerequisites

We recommend to use conda or miniconda to install the environments. The install comments are listed as following:

# Create an enviroment
conda create --name test_rdhei python=3.8
# Activate the new enviroment
conda activate test_rdhei
# Install relate packages
conda install argparse
python -m pip install scikit-image # will install nump
pip install pandas # For testing the auto.py file
pip install tdqm # For testing the auto.py file
conda install -c conda-forge matplotlib # For visualization
pip install pathlib

But if you prefer to use pip install the required enviroment for this project, there is also a file named "requirements.txt" to help.

# An Example
# Use conda/miniconda to create the same environment to run this project
# Create the environment from the environment.yml file

conda env create -f environment.yml

Running the tests

There are three files under the source_code dictionary for testing purpose: auto.py, demo.py, eval.py

Run the demo.py file

This is a demo for using EMR or LMR to test a specific image:

cd source_code
python demo.py <method> <image_name>

# for example, we test image lena use EMR method:
python demo.py EMR lena

The demo result is stored in the outputs dictionary, and the file name is "EMR_demo_lena.png".

The demo tesing result of image lena with the EMR method:

EMR_demo_lena

# for example, we test image lena use LMR method:
python demo.py LMR lena

The demo result is stored in the outputs dictionary, and the file name is "LMR_demo_lena.png".

The demo tesing result of image lena with the LMR method: LMR_demo_lena

Run the eval.py file

This is a evaluation for using EMR or LMR to test a specific image:

cd source_code
python eval.py <method> <image_name>

# for example, we test image lena use EMR method:
python eval.py EMR lena

The evaluation tesing result of image lena with the EMR method:

----- Test the image lena -----

----- Maximum Data Embedding Rate -----
The maximum data embedding rate (DER) is: 2.65667724609375

----- Secret Information Extraction Phase -----
Is it error-free? Answer: True

----- Shannon Entropy Results -----
The Shannon Entropy of the original image is: 7.44556757034006
The Shannon Entropy of the encrypted image is: 7.999250158874972
The Shannon Entropy of the marked encrypted image is: 7.919896506351618

----- Chi Square Results -----
The chisquare of the original image is: 340.6469861023652
The chisquare of the encrypted image is: 16.507810651325027
The chisquare of the marked encrypted image is: 169.09512739690342

----- NPCR Results -----
The NPCR between original image and encrypted image is: 0.9960899353027344
The NPCR between original image and marked encrypted image is: 0.9959945678710938

----- UACI Results -----
The UACI between original image and encrypted image is: 0.28583332136565565
The UACI between original image and marked encrypted image is: 0.2866824729769837

----- PSNR Results -----
The Peak Signal-to-Noise Ratio between original image and encrypted image is: 9.239278996208478
The Peak Signal-to-Noise Ratio between original image and marked encrypted image is: 9.222844749731657
The Peak Signal-to-Noise Ratio between original image and recovered image is: 51.143920872280816

----- SSIM Results -----
The Structural SIMilarity between original image and recovered image is: 0.996270899634543
# for example, we test image lena use EMR method:
python eval.py LMR lena

The evaluation tesing result of image lena with the LMR method:

----- Test the image lena -----

----- Maximum Data Embedding Rate -----
The maximum data embedding rate (DER) is: 1.9925079345703125

----- Secret Information Extraction Phase -----
Is it error-free? Answer: True

----- Shannon Entropy Results -----
The Shannon Entropy of the original image is: 7.44556757034006
The Shannon Entropy of the encrypted image is: 7.999194930860796
The Shannon Entropy of the marked encrypted image is: 7.999162516610978

----- Chi Square Results -----
The chisquare of the original image is: 340.6469861023652
The chisquare of the encrypted image is: 17.095035372148253
The chisquare of the marked encrypted image is: 17.422428522596956

----- NPCR Results -----
The NPCR between original image and encrypted image is: 0.99609375
The NPCR between original image and marked encrypted image is: 0.9959564208984375

----- UACI Results -----
The UACI between original image and encrypted image is: 0.2866433237113205
The UACI between original image and marked encrypted image is: 0.2861614750880821

----- PSNR Results -----
The Peak Signal-to-Noise Ratio between original image and encrypted image is: 9.218278226235427
The Peak Signal-to-Noise Ratio between original image and marked encrypted image is: 9.235855003727162
The Peak Signal-to-Noise Ratio between original image and recovered image is: inf

----- SSIM Results -----
The Structural SIMilarity between original image and recovered image is: 1.0

Run the auto.py file

This file is for automated testing purpose:

cd source_code
python auto.py <method> <handle> <images> # <method>: emr/lmr, <handle>: test/open, <images>: how many images will be tested/opened.

We stored the auto testing result of 10,000 imaegs using EMR method in the outputs dictionary, and the file name is "EMR_10000.csv".

# for example, if we want to open the EMR_10000.csv file

python auto.py emr open 10000

The auto tesing result of image lena with the EMR method:

################### DER Analysis ###################


----- The Maximum DER is: ----
6.268180847167969

----- The Minimal DER is: ----
1.2087364196777344

----- The Average DER is: ----
3.2456654582977293


################### MSB Analysis ###################


----- The Maximum MSB is: ----
7

----- The Minimal MSB is: ----
2

----- The Most Frequent MSB is: ----
5


################### PSNR Analysis ###################


----- The Maximum PSNR is: ----
51.17239515656845

----- The Minimal PSNR is: ----
51.110891251462775

----- The Average PSNR is: ----
51.1409200454349


################### SSIM Analysis ###################


----- The Maximum SSIM is: ----
0.9997702176548636

----- The Minimal SSIM is: ----
0.974620836378378

----- The Average SSIM is: ----
0.995820349480765

We stored the auto testing result of 10,000 imaegs using LMR method in the outputs dictionary, and the file name is "LMR_10000.csv".

# for example, if we want to open the LMR_10000.csv file

python auto.py lmr open 10000

The auto tesing result of image lena with the LMR method:

################### DER Analysis ###################


----- The Maximum DER is: ----
6.016880035400391

----- The Minimal DER is: ----
0.21290206909179688

----- The Average DER is: ----
2.532548918533325


################### MSB Analysis ###################


----- The Maximum MSB is: ----
8

----- The Minimal MSB is: ----
2

----- The Most Frequent MSB is: ----
6


################### PSNR Analysis ###################


----- The Maximum PSNR is: ----
inf

----- The Minimal PSNR is: ----
inf

----- The Average PSNR is: ----
inf

################### SSIM Analysis ###################


----- The Maximum SSIM is: ----
1

----- The Minimal SSIM is: ----
1

----- The Average SSIM is: ----
1

Authors

  • Yike Zhang - Initial work - GitHub

License

This project is licensed under the MIT License.