←back to thread

586 points prawn | 2 comments | | HN request time: 0.405s | source
Show context
schoen ◴[] No.14502425[source]
I wrote this article/originally created this list, and I would like to emphasize that there is a second generation of this technology that probably uses dithering parameters or something of that sort, and that does not produce visible dots but still creates a tracking code. We don't know the details but we do know that some companies told governments that they were going to do this, and that some newer printers from companies that the government agencies said were onboard with forensic marking no longer print yellow dots.

That makes me think that it may have been a mistake to create this list in the first place, because the main practical use of the list would be to help people buy color laser printers that don't do forensic tracking, yet it's not clear that any such printers are actually commercially available.

replies(8): >>14502841 #>>14503474 #>>14504327 #>>14504357 #>>14504856 #>>14505064 #>>14505539 #>>14507194 #
captainmuon ◴[] No.14504357[source]
Is somebody working on identifying these modern watermarks? A start would be to print out test pages and compare high resolution scans. Maybe also multiple printouts from the same printer to see what the natural variation is, and if there is a timestamp component.

I would start, but I'm currently not around a printer...

replies(3): >>14504471 #>>14504803 #>>14505308 #
RegW ◴[] No.14505308[source]
I suppose the approach is to create a machine learning dataset that maps hi-res scans of sample documents to the printers that produced them. If the resulting classifier can accurately id the printer, you have probably found a watermark, but it might just be natural variations in the manufacturing.
replies(2): >>14505531 #>>14508048 #
1. Paul-ish ◴[] No.14508048[source]
The difficulty in this approach is that you have an extremely large number of classes. Each printer is its own class. Typically, as the number of classes goes up, accuracy goes done. That isn't to say it isn't possible, but it would require a lot of custom hacks to any learning algorithm.

Also to convince anyone that it works, you would need to test it out on an extremely large number of printers, including ones of the same model. In practice that could be expensive.

replies(1): >>14508255 #
2. kpil ◴[] No.14508255[source]
Nah, it's not feasible to know the printer model if you want to identify a laserprinted dollar.

A few variants at most.