According to the European Central Bank (ECB), police, retailers and banks withdrew 55,159 counterfeit euro bills from circulation in Germany in 2019. Anyone who has once accepted counterfeit money can no longer have this refunded in real money. To prevent counterfeiting as best as possible, banknotes have certain security features. There are more than 100 different security features worldwide, but only some of them are publicly known. They help to distinguish counterfeit money beyond doubt. The "feel-see-tilt" test has proved its worth over the years. The use of special printing processes gives the banknotes their unique surface texture, which can be felt. In addition, certain picture elements such as a security thread or the portrait window can be seen, and by tilting, the emerald number on the bill changes color. Although printing techniques for banknotes are constantly evolving, experience shows that counterfeiters adapt quickly to technical innovations. This is a topic that is also being researched at the Institute for Industrial Information Technology (inIT).
For ten years, the inIT scientists in the "Discrete Systems" working group, headed by Institute Director Professor Volker Lohweg, have been working on the production and quality assurance of banknotes, wear and tear, and security at ATMs.
At the Optical Document Security Conference (ODS), the world's most important and largest conference for the field of document security, in San Francisco, Professor Lohweg presented his paper on the creation of so-called intaglio security features for banknotes, a world first. "We are showing for the first time in the world a new method, based on artificial intelligence (AI), for the rapid creation of Intaglio line patterns for banknotes," confirmed Lohweg.
"Intaglio lines, which we all know from portraits or other motifs on banknotes, used to be engraved in copper by engravers and then duplicated - a process that can take up to three months. For some time now, motifs have been created by graphics programs whose data then control laser systems that replace engraving, but this process can take just as many weeks because the engraver has to develop the motif on the computer. Our process trains an AI-which can take up to two days-but then creates a portrait in minutes. The engraver is of course still needed to select the motifs to be learned, but the machine takes over the strenuous work of creating the details. Initial results are more than promising, yet some research is certainly still needed to teach an AI an art style," reports the recognized document security expert.
Even if it still takes some time and work to implement the process holistically, the foundation for faster creation of a security feature has been laid and one is once again one step ahead of the counterfeiters.