CryptoStego

JS library for steganography with encryption - Hide text in an image with encryption and obfuscation using neural networks!

Note: If you want to use LSB mode (almost equal to original image, but is not robust to any distortion / compression) or DCT mode (rule-based algorithm that has robustness towards some image distortions / compressions), use v1.8

Version

v3.0

DEMO

http://stego.js.org Note: Library needs HTML5 support!

Test image

PNG - 635 KB

Test data

你好，世界！
HELLO WORLD!
¡HOLA MUNDO!
مرحبا بالعالم!
BONJOUR LE MONDE!
こんにちは世界！
ПРИВЕТ МИР!

Use model optimized for visual similarity

JPG compressed to 120 KB (18.9%). This is the maximum compression that test data can be decoded correctly for this model.

Use model optimized for robustness

JPG compressed to 28 KB (4.4%). This is the maximum compression that test data can be decoded correctly for this model.

You can try to decode data yourself in the demo webpage. Use empty password.

How to use

Download cryptostego.zip

Replace models in dependencies/models with the model version you want (different models can be downloaded from GitHub release). Please note, encoder and decoder must be used in pairs.

Note: This JS library needs HTML5 support!

put stego.js at same level of your index.html (together with dependencies folder) and import it in your HTML. If your file structure is different, you might need to modify stego.js

<script type="module">
    import { initCodecs, isCodecsReady, loadIMGtoCanvas, writeMsgToCanvas, readMsgFromCanvas } from './stego.js';

    ...
</script>

Features

• new in 3.0: robust to image resize (upsampling, downsampling), color jittering, photo editing etc.
• robust to compression (JPEG, PNG, GIF).
• new in 3.0: encoded image is more visually equivalent to original image.

Usage

This library provides 5 wrapper functions. Use <script src="cryptostego.min.js"></script> in your HTML to include this library.

async function initCodecs()

Call this before any other functions.

function isCodecsReady()

Tells if encoder / decoder is ready. If you initialized initCodecs but didn't wait for it, you may want to use this function.

async function loadIMGtoCanvas(inputid, canvasid, maxsize = 0)

Load image from inputid (e.g. input[type=file]) to canvasid (e.g. <canvas>) and resize it to maxsize (in pixels). If maxsize is not specified, it will be the image size. Recommended value is 600. The image should be at least 300 * 300 to be encoded properly.

async function writeMsgToCanvas(canvasid, msg, password = '', check_valid = true)

Write message msg to canvas canvasid. msg is a string. password is a string. If password is not specified, the message will be encrypted with password ''. check_valid is a boolean. If check_valid is true, the function will check if the encoded image can be decoded.

This function returns a boolean of whether the encoded image is valid. If check_valid is set to false, it will always return true.

async function readMsgFromCanvas(canvasid, password = '')

Read message from canvas canvasid. password is a string. If password is not specified, the message will be decrypted with password ''.

Minimum Coding Example

Refer to example/ folder. Or for demo page, refer to the gh-pages branch.

Copyright

Jeffery Zhao

License: GNU AGPL v3.0 or later for both code and model. For commercial use requiring another license, contact me.