I don't understand why everyone seems obsessed with complex, automated, technical solutions when a simple manual procedure will do. Or are we talking about thousands of pages (sorry, haven't read the article, just responding to the comments)?
If I were the NSA, I'd have a modified graphics driver which overlays pseudorandom very faint grey dots over the screen at all times. A 254 254 254 pixel hidden amongst all while pixels isn't visible, yet thousands of them across a page will encode significant amounts of information, even in the face of quite severe image compression and low quality.
The dots could be based on the computer, currently logged in user, and timestamp.
Then later, if any screenshot or screen photo is leaked, you can decode the dots to identify the source.
You think that they would even be picked up when you take a picture with a camera? Between the external camera, and then compression, I don't think that the 254 254 254 pixels are going to make it into the final image. They might not even make it into the initial picture - screen backlighting consistency, etc, is going to wreak havoc on that from the start, before we even get into sensor noise, etc on the camera, any smudges on the lens, all before it even gets saved a jpg
Which shade of yellow? Assuming it's a uniform yellow (which I doubt in an analog to digital conversion), miss it by one bit, and your source is burned.
Sure. Quick thought experiment: you've got a 24-bit scanner, which means 2^24 individual values, so 16.8 million. Let's say 2 million of those are yellow. How many different shades of yellow are you going to get from your scan? Do you take the average? Does the process leave artifacts - that is, does it print yellow over yellow and leave a physical trace? My point is you've got to make sure the signal is well below the noise floor because we're talking about the NSA, and there are more sure ways to filter this signal.
Just OCR it.
Or just retype it.
I don't understand why everyone seems obsessed with complex, automated, technical solutions when a simple manual procedure will do. Or are we talking about thousands of pages (sorry, haven't read the article, just responding to the comments)?
dont print, make a picture of the screen with old camera bought in a car sale town away.
If I were the NSA, I'd have a modified graphics driver which overlays pseudorandom very faint grey dots over the screen at all times. A 254 254 254 pixel hidden amongst all while pixels isn't visible, yet thousands of them across a page will encode significant amounts of information, even in the face of quite severe image compression and low quality.
The dots could be based on the computer, currently logged in user, and timestamp.
Then later, if any screenshot or screen photo is leaked, you can decode the dots to identify the source.
You think that they would even be picked up when you take a picture with a camera? Between the external camera, and then compression, I don't think that the 254 254 254 pixels are going to make it into the final image. They might not even make it into the initial picture - screen backlighting consistency, etc, is going to wreak havoc on that from the start, before we even get into sensor noise, etc on the camera, any smudges on the lens, all before it even gets saved a jpg
1 reply →
Which shade of yellow? Assuming it's a uniform yellow (which I doubt in an analog to digital conversion), miss it by one bit, and your source is burned.
Well... You scan your printed page in and scan the dots and see what colour they are, then you change the background to that colour...
Sure. Quick thought experiment: you've got a 24-bit scanner, which means 2^24 individual values, so 16.8 million. Let's say 2 million of those are yellow. How many different shades of yellow are you going to get from your scan? Do you take the average? Does the process leave artifacts - that is, does it print yellow over yellow and leave a physical trace? My point is you've got to make sure the signal is well below the noise floor because we're talking about the NSA, and there are more sure ways to filter this signal.
1 reply →