Slashdot Mirror
New Optical Security Doesn't Require Embedment
An anonymous reader writes "Scientists are claiming to have a new type of optical security that doesn't require embedment. Optical security includes many different options but up until now they have all required that the secret image be embedded in a host image which left it vulnerable. From the article: 'To address this problem of finding the secret image in the watermark, scientists have developed a new optical security method that doesn't require embedment. Instead, the technique uses a phase retrieval algorithm to generate specific optical and phase keys that extract the secret information when applied. The optical keys contain information and are distributed to an individual through a personal identification number (PIN). The information contained in the phase keys (the main source for determining extraction) is distributed to the individual separately.'"
I even RTFA'ed, and I still don't understand exactly how this technique is new, or even useful. Someone care to clarify?
Oh, and Google says "embedment" doesn't mean quite what you think it means. :P
-:sigma.SB
WARN
THERE IS ANOTHER SYSTEM
A fine cromulent word with which to embiggen your article.
I think I speak for most of us who has only read the summary when I say: huh?
Bjarke Roune
i want enlightenment on the article first.
is the summary in greek or in latin?
fifteen jugglers, five believers
Ok, so I read the article. Not much clue there. I thought to start with that they were talking about some kind of steganography, but the article claims that the encoded message isn't embedded in the image. It's not a digital image either, (jpeg etc) it's a printed image. As far as I can tell they're using some optical properties of the image as a key to decode some other encrypted data. Hardly an earth shattering technique, but the linked article is just a brief, confusing write-up of an optical physics paper - perhaps there's actually something interesting in the paper that got dropped along the way.
.evom ton seod gis eht
Storing information by modulating the Fourier (or Fresnel) modes of an image is not new.
That being said, the actual underlying science of this post might be intersting, if only I could get to it through the torrent of drivel in the summary.
e.g. "meaning that the secret image cannot be found in the watermarked image"
Then how do you extract it then?
Do you mean "the image cannot be extracted without the key"?
"Since the watermarked image contains no secret information" Que?
If i understand this correctly, traditionally you use steganography to hide the secret (another image) inside an existing (host) image, with a key to decrypt it, the draw back being that someone might accidentally spot the steganography.
This technique doesnt put any data in the host image at all, the keys contain all information required to distort the host image into the secret image, thus given the host image, you cant accidentally stumble across the secret, and likewise the keys are of no use unless you also have the host image.
Its akin to having a text encryption system where the key is the offsets into a known document where the letters can be found, the known document can be public, but unless you know what both it and the key neither is of any use.
Maybe an interesting mod for this guy. I'd say informative, but after RTFA I still can't be certain he's got it right. It's a convincing summary though (he'd win my vote in a game of dictionary).
Perhaps there's a lot more to this than I can reasonbly imagine, but wouldn't the keys need to be almost as large as the original image, unless the original is closely optically related? Why not just diff the images, encrypt it, and give the person who enters the PIN properly the diff and the key? There's no mention of how large the keys are, so one can only assume that this is news because the keys they're talking about are significanly smaller than a compressed diff of the image. Or not.
Is it just my observation, or are there way too many stupid people in the world?
Once he was embedment he was enbiggened and she was...
Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong fix.
a body of text is publically distributed, and a key is distributed which modifies the text into some secret message?
you could use a piece of known text too, like the gutenberg projects first copy of the book of genesis.
or you could use a changing piece of text, like the html of news.google.com, this would make the key automatically useless after a certain time.
Of course you don't see it. The message is hidden in the least significant bit of every letter.
404555974007725459910684486621289147856453481154 in hex is "You sank my Battleship?"
[GPG key in journal]
When the author says "doesn't require embedment" it means that the "host image" is not altered in any way. Since there is no information provided to potential hackers through discovery of an invisible watermark it is far more robust than traditiona stenographic techniques.
I gather the way it works is that the "hidden image" is like the "secret key" you use when you generate SSL certificate requests--it is kept secure/inaccessible to the public. Through transforms and other mathematical wizardry they calculate optical/phase keys--essentially parameters you'd feed into the transform functions. This is the "public key" as you'd use in SSL and is expressed as a "PIN".
The way the security would work, the user being authenticated would have to provide the "host image" and the "public key/PIN" to prove they are the right person. If you feed the parameters from the PIN into the transform functions and run that function over the host image, the result would match the "hiddem image" (secret key). If either the wrong image or wrong PIN are supplied, the calculated image would look distorted and authentication would fail. Essentially you are creating a mathematical function that would (in the article's example) warp the picture of the pagota into the picture of a cute chick, and the PIN provides the missing variables in the function.
So...can someone out there confirm that this is the idea? That's the gist I got from it anyways...
Thanks, Aceticon. Next time you get to write TFA and TFS. :)
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
If I understand this right, we have some base image, Image, that is non-secret. And then we have some function to derive our "decryption key" of sorts, ModulatePhase(Image, Secret) => Key, and to decrypt, we use the inverse function InversePhase(Image, Key) => Secret?
:-) This seems sorta like the "key files" (using random files on your HD as part of your 'password' to encrypt & decrypt) in TrueCrypt.
So, if we make ModulatePhase == InversePhase == XOR, and 'Image' be any binary file of sufficient length... umm. Yeah, we might reveal the message length unless there are good ways to figure out the original length of the file (otherwise, we know that the length of the key is the length of the file) and probably other practical difficulties, but the principle isn't so different... Although maybe the extra fancy mathematics does something for such issues.
Umm, so, what was the point of using an image again?