3 ms·
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 proje
by ChuckMcM 11y ago
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.
- zeveb 11y ago> 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. Wouldn't encrypting the data to be hidden, and then hiding it within random data (say, the low bits of image bytes) be perfectly indetectable? I'm assuming that the low bits of image bytes are uniformly random, but even if they're not then one could still hide data in them, it's just trickier.
- ChuckMcM 11y agoThat was the point of the Citi paper[1], to test whether or not you could detect the presence of hidden data in images. The did so by analyzing the statistical properties of the images. [1] http://www.citi.umich.edu/u/provos/papers/detecting.pdf http://www.citi.umich.edu/u/provos/papers/detecting.pdf
- minimaxir 11y agoThe goal of the post was "brilliantly horrible" so it appears I have succeeded. :p Another issue I am looking into is "could an image steganography SaaS startup exist today?" I believe the answer is no because web services are inconsistent about how they modify the picture in transit. (e.g. Twitter lossly compresses the photo with no way to retrieve the original.)
- rzzzt 11y agoDigital watermarking methods are usually resilient against some levels and kinds of manipulation of the original image (flipping horizontally or vertically, cropping, magnification); JPEG compression can probably be accounted for as well.