Video steganography

[chants]

Does anyone know of any practical projects to embed data into a video that could be recovered in a method similar to scanning a QR code, with the reliability and checksum features provided by QR but that would also be difficult to near impossible to detect besides using a special application that could decode it in a lossy state from for example a phone camera. That would be pretty neat to say the least.

[atom0s]

There’s several projects in the past that abused ‘fake’ videos to use YouTube as file storage. The issue now with that setup is that YouTube compresses data causing losses. But the projects themselves still exist and can be used to create the needed video files.

You can find various projects on Github to do it:
https://github.com/search?q=youtube%20file%20storage&type=repositories

As for another project idea, which I would assume this also already exists, but you could create a tool that takes in a file/data and splits it into chunks of data that would fit into QR codes. Then have the tool generate images of each of the QR codes in order, and then stream those to ffmpeg or some other video library to generate a 1-frame-per-image slideshow video of each QR code. The end result would be the file/data encoded into QR codes stored in a video that is 1 frame per chunk.

You could then make another tool that reads the video data back frame by frame and automatically convers the QR code back to that given chunk of data, do that for each frame and you can rebuild the original data. Assuming no losses in frames it should be able to convert back to the original file/data.

[chants]

Yes this idea sounds the most practical with a QR frame sequence. Technically cracks or keygens could have their binary data distributed in the video. And if that data were encrypted with AES whose key was a hashed password then it would be secure even. Though it's not stealthy in that it is obvious data is hidden in the video. I was thinking some color channel manipulation to do the QR might be possible. After all, something like gridlines with slightly faded colors would be highly visible to a computer vision algorithm but nearly imperceptible to a human.

QR code sequence with minimum frame holding to ensure lowered frame rates don't cause an issue. Plus using color channels in a clever way instead of black and white. Not to mention some encryption just in case. That's an interesting encoding and decoding tool to have for sure.