Unlimited Data Storage Using Image Steganography and Cat GIFs

That's pretty neat, but it can be way easier. File formats have never really been robustly verified. The difference between `file` telling you that a binary file is "data" versus your image format is about 3 bytes.

I wrote a program[0] a few years ago that does exactly this "steganography" without encoding it in as pixels. It just prints a tiny png and then concatenates your data on to the back of it. Even my script could be way simpler if I was a better programmer at the time =]. To image sites, etc it's all the same. I thought it was pretty cool when I did it. Simple concept. Worked well. All you need is to encrypt your data first, since stego is hand wavy mumbo jumbo. Here[1] is an example

[0] - https://github.com/ejcx/pnger/blob/master/pnger.py

[1] - http://i.imgur.com/J9OJZfS.png

>The potential amount of encodable data is larger than the file size of the image itself! Screw secret messages, we’re literally creating data storage out of nothing!

Uh, nope. The file is small because the frames are non random, and thus heavily compressed. If you wanted to store arbitrary data, the frames would be far more random, and therefore take up more space.

(The above was written halfway through the article)

I disagree that you need "deep knowledge of how image compression works". I have very little idea how the actual compression works, the only thing I needed to know was that images are compressed and random/arbitrary data doesn't compress as well for fundamental algorithmic reasons. It's obvious that changing images can make them larger even without any knowledge of the particular compression used.

Can this be demonstrated as true? I'm pretty sure the only reason it appears that the cat GIF could contain itself is because of compression. I imagine there's some mathematical proof that by adding more information effectively, you'll likely reduce the compression ratio and thus this infinite "trick" doesn't hold up. Happy to be proved wrong.

I'm on the fence if this is horrible or brilliant. On the one hand the technique has been available for decades, and UMich's Citi group did an interesting project to see if people could both exchange messages on Ebay using pictures of things for sale, and if so did any images on Ebay have steganographic content (at the time it was the largest source of user supplied images). They didn't find any (except for their own test images). But it presumably continues to be a channel people check.

Now you flood the web with images that have had data added to them, and it creates potentially millions of false positives. That then actually enables using the channel for covert conversations because so many images are now carrying data, they no longer stand out.

Anyway, given that a 2TB drive is like $90, so you can create a dual parity raid set of 6 for less than $600, that seems like way more storage than you'll probably use and if you amortize it over 3 years its way less than the cost of your Internet connection over that same time.

When you understand basic information theory you know this is impossible.

In gifs you have a separate "color map" that stores at most 256 colors. So by modifying the LSB of each color you get at most 256 bits aka 32 bytes.

Many years ago I was asked how to convert a .gif file to a .wav file. I remember taking the time to patiently explain that it just doesn't work that way. Looks like I owe someone an apology.

Neat trick. I'll pass on dropping tens of millions on your startup. Might still use the trick for storage, though. I'll run the stuff through GPG or bcrypt first. :)

Except that if your adversary finds your secret trove of a billion cat GIFs somewhere... then it isn't "steganography" anymore.