[image of digits]
Steganography - Currency Detection Algorithms

Wired: "Currency Detector Easy to Defeat," by Chris Ulbrich.
Freedom to Tinker: "Photoshop and Currency." by Ed Felten.
Digital Media: On How to Defeat PhotoShop.

Questions? Comments? Please email me... nick.gessler(at)duke.edu

All images are in compliance with the the Central Bank Counterfeit Deterrence Group (CBCDG).


 

Photo reproduced courtesyof Steven J. Murdoch.
"Software Detection of Currency," by Steven J. Murdoch.


Image source: U.S. Bureau of Engraving & Printing.

Adobe PhotoShop CS Currency Recognition Tests:

Exactly how is currency recognized?
These tests were run on larger 3678 x 1560 pixel images. Thumbnails of the test images are shown below:
Not Recognized as Currency:
Recognized as Currency::
x
The pastel background has been whitened out.
x
The image has been flipped horizontally.
x
The image colors have been inverted.
  
x
Only the pastel background has been blackened out.

x
The portrait also has been blackened out.
x
The "United States of America" text also has been blackened out.
x
All text and features in black printing have been blackened out
x
The blacks and grays have been whitened out.
x
The blacks and grays have been whitened out and then the colors have been saturated.
x
The blacks and grays have been whitened out, the colors have then been saturated and the levels adjusted.
x
The blacks and grays have been whitened out, the colors have then been saturated, the levels adjusted and the image has been changed to grayscale.
x
Currency Not Recognized. Everything but the pastel borders has been blackened out.



x
x
 x
x		  


A segment of the hexagonal background pattern has been tiled to the size of the banknote image.


Preliminary Conclusion: The key to recognition is in the pastel background. But what is the algorithm that is being used?