IDs in Color Copies
Slashdot received a lot of submissions of this Privacy Forum article about ID numbers being "watermarked" (just like digital watermarks) into copies made by any color copy machine. Go ahead and read it; the rest of this story assumes that you've read the link.
This not a secret; I remember a case a few years ago where a Columbia University copier was being used to create counterfeit currency, and the imprinted copies were traced straight back to the machine used to create them (amazingly, Altavista turned up an article about this case). Basically, when color copiers first started getting good, the Treasury started leaning on manufacturers to make their products less useful for counterfeiting. AFAIK, there's no law in effect saying that manufacturers MUST include an anti-counterfeiting features in their devices; but on the other hand, there aren't very many equipment manufacturers, so they're easy to lean on.
So today, any copy you make with any color copier will include a unique serial number. Make sure you don't copy anything that someone might want to trace back to you on a color copier. Maybe this isn't that big a deal; color copiers aren't home appliances.
But now home scanners and inkjets make up a nice copying system for as little $200-300. The Treasury Department has a big program devoted to preventing digital copying, and it looks like one of their main concerns is consumer-grade equipment. The Bureau of Printing and Engraving is even soliciting proposals from vendors which have a system suitable for embedding these watermarks in all output produced by color inkjet printers.
Fighting counterfeiting is fine with me. Thus the systems which "recognize" currency and refuse to scan or print it don't seem like too much of an infringement. But embedding serial numbers in all printer output? Maybe I just have a cynical mind, but I can think of about a hundred reasons this is a bad idea.
/.
/. If the government wants us to respect the law, it should set a better example.
Canon color lasers (800, 1000, 2400, ect.) all have a board that recognizes things like money and postage stamps. If you try and copy any of these it will spit out all black copies, and will continue to do so until a Canon tech is called. (They usualy call the Secret Service)
Paper and print is a whole different story. I would be wary of buying anything that watermarked everything I printed. I use a (granddaddy) AppleWriter II laser printer, and am reluctant to upgrade to a new printer if I am aware of this sort of thing. My big concern is who can read these watermarks, and why would they ever want to. (Other than for legal reasons, but I can't see myself printing threats off my printer.)
I can see newer laser printers being able to do this sort of thing, but I cannot see why a printer company would risk the public relations disaster that would ensue after someone found it producing a watermark, and any possible corporate backlash from including such a "feature".
I really don't think that much about my privacy, I'd like to think I'm pretty good to get along with in that way, but I (personally -- someone will hopefully point out valid reasons, but I guess valid reasons depend on who can read the watermark ...) can't see any justifiable reason for said watermarks except for perhaps malicious purposes.
The article says that we may see this sort of thing implemented in ink jets soon; I'm hanging onto mine.
Ink jets have come a long way in the last few years, and they've reached the stage where, with the right paper, they can print photographic-quality pictures.
What does this mean? Well, everyone who's planning on doing something nasty-and-traceable will do it on an older printer. Some stupid people won't, and they'll get caught, justifying in the minds of the Man and the public that such watermarking is worthwhile. But, like drug smuggling, the vast majority will slip by unnoticed.
Freedoms will be curtailed, money will be wasted, and it'll all be for nothing. Have a nice day.
Privacy concerns aside, the only thing about placing an identification number on a color print that would bother me would be if I could see it. If the ID was scattered about the page as "noise", and unobservable by me, it wouldn't bother me much.
As for the privacy issue... wouldn't such a encoding method be proprietary to the manufacturer? So what happens if I first copy the color image on a Xerox machine, and then take the copy over to a different machine, and copy that. Assuming the quality was not lost, the hidden ID code would not be decipherable by any (one) decryption algorythm.
Mike Eckardt meckardt@spam.yahoo.com
I saw some earlier concerns about the tracability of a particular printer... The watermark contains the serial number of the printer. Previous techniques of forensic science already allow us to identify printer makes and models, so the change in watermarking will not assist law enforcment at all (aside from possibly knowing which store the device was sold at), as long as you just don't fill out that product registration card...
Which no counterfeiter would do anyways.
So why bother at all? It will make printers more expensive, and the government thinks that they get a tool to assist them in enforcing the law, but doesn't really - other coroberative evidence will have to be collected to get near enough to the printer with the watermark to check, and then traditional forensic science techniques could be used.
Don't like my sig? I don't either.
Sounds like a great way to make the lives of the people at Kinkos a living hell. Find a print sample that sets off the board and incorporate it into your letter head... Then just lurk in at odd times and leave a trail of black-copying color copiers in your wake...
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
A second alternative is based on the premise that such an image is necessarily going to be faint, to prevent it obscuring what you're printing. If this is to scale with your print-out, then simply print your image much fainter (making the watermark effectively invisible), before reprinting the image upside-down on the same page, with the paper also inverted. This'll put a second copy on the paper, making it normal-strength, but the watermarks will only overlap in places (one being 180' to the other), so rendering most of it invisible.
Finally, switch back to a daisy-wheel. I don't care =HOW= good a manufacturer is, they can't make a daisy-wheel print watermarks, come hell or high water. Besides, daisy-wheels are great for listing print-outs. That is, if you want to turn the recipient into a gibbering idiot. :)
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
If anyone has ever encountered a digital watermark of the Digimarc kind, (Photoshop users know what I'm talking about) there's an easy way to remove it. Resize your image to 95% of the size then resize back to 100% - minimal loss of quality (especially at high resolutions) and no watermark.
This is just an example but Digimarc underlines two serious problems with watermarks:
a) No watermark is invisible. No matter what anyone tells you adding watermark is a lossy process. The harder the watermark is to remove the more visible it is.
b) Watermarks will always be removable. If you have physical access to the machine creating the watermark (scanner, printer, whatever) you'll be able to edit/disable the watermark. Unless you embed the watermark in the paper fiber or something (but then you're tracking the media, not the data) it'll probably be enough to cut off part of the image or something similar to disable the watermark.
jay
... and this is something we've known about for years. Most color copiers do embed a serial number and many--particularly canons--will shutdown if you try to copy currency.
While the owner of the copier may not be officially required to register with the manufacturer, most non-consumer grade equipment needs to be serviced at least once a month. For example, each color copy generates a small amount of excess toner which is scraped off into a waste toner bottle; Xerox decided not to make this a user serviceable part on the Docucolor 40's (which are in almost every Kinko's in the world).
Kinko's, however, is generally more interested in making money and avoiding lawsuits than invading anyone's privacy. Every Kinko's Co-Worker is trained in the copy guidelines generated by our pack or ravenous lawyers about what we can copy and how. For example, the kid in the article should of been told that we can copy his driver's license but only in black and white and only at 129 percent.
Anyhow, for your extra dose of paranoia today consider this: even most of the new black and white copiers (from the Docutech to the Xerox 265) actually digitize and and store the images rather than flashing them to an analog transfer belt. All these copiers are equiped with a modem.
In Indiana it is illegal to make a monkey smoke a cigarette.
I really don't understand the point of getting excited about this. The police can probably already match paper and ink, and minute impressions in the paper from handling to identify a specific printer.
The things you mention can be used to make a positive identification of the printer...but they have to find it first. The serial numbers, however, can be looked up in a manufacturers database. For those that didn't read the article (I almost didn't but I'm glad I did) here's something interesting from it:
To read these IDs, the document in question is scanned and the "noise"
decoded via a secret and proprietary algorithm. In the case of
Xerox-manufactured equipment, only Xerox has the means to do this, and they
require a court order to do so (except for some specific government
agencies, for whom they no longer require court authorizations). I'm told
that the number of requests Xerox receives for this service is on the order
of a couple a week from within the U.S.
In other words, according to the author, Xerox routinely, sometimes WITHOUT REQUIRING A WARRANT, gives out information regarding ownership of copiers based on these ID numbers.
numb
Second, consider this problem from the point of view of the imaging equipment manufacturers. They clearly do not want to have some ill-conceived legislation rammed down their throats, that mandates some expensive technology be put into every device built. Even worse would be if every nation they sold too had different regulation that needed to be complied with.
So, the manufacturers with foresight are cooperating with the government to try to come up with inexpensive solutions that make the government happy.
Now, keep in mind that anticounterfeiting is distinct from encoding distinct marks into documents. I'm going to talk about the former first, and will come back to the latter.
As the original article pointed out, there are already both anti-counterfeiting and source indentification features built into current color copiers. However, these solutions are not necessarily extendable to consumer imaging, because they take place in a closed architecture as opposed to an open one.
Consider the path that a counterfeit note would take in a consumer based imaging system. It would travel from a scanner, to some image processing software, to the printer. At any point along the way it could be stored, manipulated, or transmitted via the internet. Each step may or may not be carried out by the same individuals. Becasue of this, the most logical place to put counterfeit prevention is in the printing step. The reason for this is that if the protection in the scanning or processing step is broken by any individual, then a print-ready file could be distributed to many others. However, putting the prevention within the printer makes it both less accessable to crackers, and requires that each potential counterfeiter break the protection again.
Let us focus then on the problem of the printing of counterfeit currency. Three ways of helping to solve this problem quickly present themselves.
1: Add currency-detecting logic to printers.
2: Add features to currency that printers cannot reproduce.
3: Add printer-specific watermarking to printers.
The trend in ink-jet printers is to make them cheaper and cheaper. Given this trend, it is not feasible to add much computational power to the printer without increasing their cost. For this reason, attempting to do general-purpose detection within the printer is not feasible at this time. This does not preclude, however, doing some quick and dirty detection that is computationally very simple. (For example, is the document being printed approximately 6" by 2.5").
The design of US currency is unsophisticated, especially when compared to that of other nations. Personally, I appreciate the simplicity and history of our design, but from an anti-counterfeiting point of view it is a nightmare. The latest iteration of our currency was a stopgap effort to try to make it somewhat more difficult to be digitally copied, but most of the new features (except perhaps for the watermark) are not well understood by the public at large.
Some simple features that could be implemented, easily recognized by the public, and impossible to duplicate on a consumer printer could include:
1: Printing on a transparant substrate.
2: (1) with areas that require perfect front-to-back registration.
3: Printing with reflective (foil)inks.
But as long as we are creating a new currency, we could consider hybrid solutions. That is, embedding some special patterns in the currency that are trivial for a scanner/printer to recognize, yet do not occur in other document types. This could be some specific geometric pattern, or a specific use of colors.
So, finally, we come to embedding a watermark in all images printed by an consumer printer. First, be aware that some office printers already do this, mostly color laser printers. But beyond that fact, this is really a separate issue from anticounterfeiting. Printer manufacturers are not going to volunteer to do this unless they are compelled to do so by governments.
But if a manufacturer were forced to implement this, the most obvious place to do so would be in the driver. This is because (as mentioned earlier) the ink-jet printer itself has little computational resources. The driver has the full resources of the system CPU and is traditionally where the dithering takes place. As far as making the patterns unique, one could either query the printer for a serial number, or simply use the Pentium III serial number, or perhaps the MAC address from the LAN card.
One closing lesson here, if you were feeling cocky by saying "no problem, I'll hang onto my old printer", then make sure you never upgrade the driver, as that is where this would most likely be implemented.
Well, in closing, I would love to sign this with my account name. I can't however, so just call me...
Anonymous Coward