This is a archived project. See http://blogs.23.nu/disLEXia/stories/492/ for details and further pointers.

Tuesday, 03. June 2003

Anti-OCR-fonts.

At the the 2nd Annual Workshop on Economics and Information Security Hal Varian was wondering why there are no substential infringements on books. Other attendees told him that there are: People are scanning and OCRing just every kind of book. Bruce Schneier was wondering, if we need anti-OCR fonts.

There are actual fonts which are in some sense Anti-OCR-fonts. They use Postscript fonts selecting characters at runtime from different other fonts. This effect is somewhat popular to produce "ransom notes" in Postscript.

The people at Letterror designed so called flipper fonts which basically do the same but use specially crafted glyphs for this process and are therefore much more readable.

Letterror also came up with the probably most useful idea of a type technology to circumvent OCR, a technique called RandomFonts, "type programmed to change in the printer so that each character would be unique" by modifying the points on the curves defining letter shapes. RandomFonts can be finely tuned to find the right trade-off between variation among characters and readability.

There are already some real-world useage of anti-ocr tehniques: Several Web-applications ask users to retype a text displayed as an distorted image. This is used to provent automated subscription. While PayPal uses a quite boring way of distortion (see above), Yahoo! deploys a wide array of possible distortions. (See Examples 1 2 3 4 5 6 7 8)

To my knowledge there has been up to now no research on the topic of anti-OCR fonts.
09:32 | #

<< insert_coin censorship experiments in english. | Government mandated blocking of foreign Web content >>

disLEXia, a research project by Maximillian Dornseif


June 2003
 
Mo Tu We Th Fr Sa Su
1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30
May

Search: