1. 2003. május 6.Digital watermark1 Alfréd Rényi Institute of Mathematics  László Csirmaz  Gyula Katona  Dezső Miklós  Tibor Nemetz HP ( Compaq )  László Marsovsky  Attila Haraszti Drótposta Ltd.

2. 2003. május 6.Digital watermark2 The problem  Current method of identifying legal or high value documents: Producing complicated patterns via printing technology (e.g. banknotes) Producing complicated patterns via printing technology (e.g. banknotes) Applying holographic markers (e.g. tax stamps) Applying holographic markers (e.g. tax stamps) Specially prepared medium (e.g. embedding metal strips into paper) Specially prepared medium (e.g. embedding metal strips into paper)  At the present time, the most common technique of authentication uses serial numbers or in general alphanumerical strings, rarely combined with pictograms. These can be copied or falsely created. We need a cheap automated solution!

3. 2003. május 6.Digital watermark3 The promise  Finding a cheap solution (mark) of identifying legal documents (bank cards, banknotes, stock certificates), which is: Unique Unique Very difficult or expensive to copy or duplicate Very difficult or expensive to copy or duplicate Can be measured and processed quickly Can be measured and processed quickly A couple of hundred bits can be generated as a check code A couple of hundred bits can be generated as a check code  Goal: An unforgeable, unique physical identifying mark An unforgeable, unique physical identifying mark A cheap device for reading the mark A cheap device for reading the mark An algorithm making a unique and short (200-500 bits long) extract from the picture An algorithm making a unique and short (200-500 bits long) extract from the picture

4. 2003. május 6.Digital watermark4 Previous achievements Different surfaces – that are actually difficult to copy, but often expensive – are measured, and the results are stored in a central database. The results of the control measurement are compared to the stored data.

5. 2003. május 6.Digital watermark5 Novelties  We do not store results  The mark is cheap, but difficult to copy  IT WORKS

6. 2003. május 6.Digital watermark6 The mark

7. 2003. május 6.Digital watermark7 The reading device

8. 2003. május 6.Digital watermark8 The method  A three-dimensional thin layer, a small mark – which is very difficult or expensive to copy or duplicate – is inseparably attached to the object. This small mark will contain randomly positioned small objects.  Using digital imaging technology we test a photograph of the object, to see that the mark is 3 dimensional, and not just a 2 dimensional copy of a 3 dimensional object.  Using different mathematical algorithms we generate a unique number (check-code), which can clearly identify the object.

9. 2003. május 6.Digital watermark9 The material Into a thin (typically 0.3-1 mm thick), translucent layer without a light reflecting coating small objects (e.g. glass beads) are embedded. Thin film 0.3-1 mm Small (5-20 micrometer) glass beads (balls) coated or not

10. 2003. május 6.Digital watermark10 Unforgeable Photos taken in diffuse and direct light

11. 2003. május 6.Digital watermark11 Results  The mark is very cheap, as we use a material available anywhere, the price could be around 2-3 cents.  The measuring device is a WEB cam with special optics, which costs 200$-300$  The stability of the ”code” is excellent. All measurement was in a tolerance of one digit. (plus, minus one pixel!)  The measurement time with non-optimized code is around 10 seconds.  We have developed a method, which does not require storing the code and checking the Hamming distance, but regenerates the code every time on line, using a small piece of redundant information, like a PIN code.