The Unofficial Apple Weblog (TUAW)

You know what's funny I made a post regarding this, in my blog a few months back. What's funny that not all Bill images trigger a warning box like the bill that I found on the internet. I have tried other currency images, no dice. Apparently some currencies get detected by the "secret" algorithm meanwhile others don't.

http://voidin.blogspot.com/2005/11/want-to-counterfeit-money-dont-use.html

Reply

↓↑report

I believe that the open-source Gimp doesn't have this "protection." I use that instead of any of the commercial utilities. There are several OS X builds linked from their site.

Given all of the legitimate reasons to want to scan and manipulate images of currency, it's pathetic that companies like Adobe and scanner/printer manufacturers agreed to put this block in their products. I always hate it when gov'ts or companies decide to interfere with legitimate users just so they can inconvenience criminals.

Reply

↓↑report

Darn it, the URL for the Gimp was stripped out. Let's try this: http://gimp.org/

Reply

↓↑report

You know after reading those slide presentations, I couldn't help but wonder if the detection code will be placed one day system-wide. In the presentation it talked about making it part of the linux kernel, for example. Although, I doubt it may happen, one can never be to sure.

Reply

↓↑report

Fabienne:

Thanks for your post. I am glad you liked my lecture at the 21C3. I should follow up on your point about colour-skewing as I think it is conflating two different anti-counterfeiting measures. Sorry if this wasn't clear from my talk.

The reason bright orange comes out as dirty brown is due to a fundamental limitation of 4-colour subtractive printing, as used by standard colour printers and photocopiers. These can only produce 3 pure colours: cyan, magenta and yellow (a separate black is used to increase the available contrast). All others are made by mixing some combination of these.

Each pigment added to the mix subtracts more wavelengths, so the more ink there is, the darker the result. This means that when a colour requires a combination of the primaries, it cannot look as vivid as those which require only one. The book "Optical Document Security" which I mention in my talk lists bright orange-red, fresh apple, emerald green, cornflower blue and pure blue-violets as being hard to reproduce with a 4-colour process. These are described as being outside the printers gamut (http://en.wikipedia.org/wiki/Gamut).

Banknotes use a spot-colour process, where a collection of pure pigments are especially chosen for a particular image. The set-up costs of this are high, so it is only suitable for material produced in large quantities. This process allows for many colours to be reproduced which are impossible with a 4-colour printer.

Banks are aware of this so deliberately choose prominent areas to be printed in a colour which is significantly distorted by a 4-colour process. This is where my example of orange becoming brown came from. More sophisticated printers use 6-colours, for example Hexachrome (http://www.pantone.com/products/products.asp?idArea=6&bShowProducts=1) which adds orange and green to the normal primaries. I have not seen how this would perform in reproducing banknotes, but I suspect it would be much better.

A separate issue is what printers do if they detect that banknotes are being printed. The "Counterfeit Detection System" which my talk discussed simply prevents the image being printed at all, but there have been rumors of previous more subtle schemes which allow a distorted version to be printed. These were perhaps proposed due to a concern that legitimate images might be falsely detected as currency.

One possible mechanism, proposed by Hewlett Packard (http://www.hpl.hp.com/news/2003/july_sept/counterfeit.html), is to detect when a colour of appropriate hue and density for currency is present. The printer would then distort it, either by changing the colour or modulate it to produce visible bands. Perhaps this is what you are thinking of. Another proposal was to tweak the position of front and back in duplex printing, which would be visible through registration marks, particularly prominent on the Euro.

Reply

↓↑report

I tried scanning 2 notes from different countries:

1st was a 10 Tolarjev bill, which is the currency of Slovakia. Worked perfectly using ScanWise and scanning directly into Photoshop CS2.

2nd was the standard 20 �-bill, with same scan programm: meep. Gave me an alert saying that Adobe doesn't support scanning of 'banking images'. I could, however, open the picture. It just notified me that I won't be able to print it.

I guess, they put into that algorithm the characteristics of the leading global currencies: the US Dollar, the �, probably the Yen as well - currencies of smaller countries are not recognized as such.

Would be, anyway, cool to find out which work and which not... .

Reply

↓↑report

Sebastian:

There was a proposal to force software providers to include the detection system in their software, but I have not heard anything recently about the progress of this law. The reason I suggested it might have to go into the Linux kernel is that it is very difficult to define what image manipulation software is.

For example, GIMP clearly is, but I also use the Python Imaging Library for modifying images, so must this include the detection code? I have even used Perl for editing PNM images, using only the text processing facilities. Detecting watermarks in images used in this way would have to be done by the operating system.

Reply

↓↑report