Concordia University researcher uses AI to identify counterfeit coins

Loonies with the the effigy of King Charles on them are struck at an event celebrating the first coin struck at the Royal Canadian Mint in Winnipeg on Tuesday, November 14, 2023. THE CANADIAN PRESS/John Woods

A researcher at Montreal's Concordia University and her team have developed an innovative technique for accurately identifying counterfeit coins using artificial intelligence (AI).

The techniques used by fraudsters to produce counterfeit coins are becoming increasingly sophisticated, sometimes confounding even the most experienced experts, but AI could change the game. Solicited by the Danish police, Maryam Sharifi Rad — a postdoctoral fellow at Concordia University's Centre for Studies in Pattern Recognition and Artificial Intelligence — and her team set about the task.

"It's a problem that very directly affects the economy and public finances," the researcher said. "I was keen to push back the limits of technology to improve the security and protection of our monetary systems."

To get started, the research team digitized a large number of genuine and counterfeit coins in high resolution. Obtaining counterfeit coins from several countries was quite a challenge, according to the researcher, since most of them had been seized by police forces around the world.

The 2D images were then analyzed with the help of technology developed by the researchers based on machine learning and image mining.

By searching each piece for irregularities and flaws often undetectable to the naked eye, the AI was able to pinpoint with great precision which pieces were genuine and which were not.

"It was really fascinating to see that AI was really able to detect extremely subtle differences between parts. Thanks to this advanced image processing method, we achieved a 99 per cent success rate," Sharifi Rad said.

Practical applications

Sharifi Rad believes that the approach developed by her team demonstrates the full potential of AI in the fight against counterfeit currency. Her partner in research and in life, Saeed Khazaee, a research associate at the centre who also contributed to the study, has also been working on a model capable of processing images of coins, this time in 3D.

Both are enthusiastic that authorities in several countries are interested in the research they have carried out, and are considering using their approach to combat fraud.

"The Royal Canadian Mint is very interested in our work, as there are a lot of counterfeit coins in circulation on the market today," Khazaee said.

In addition to the possibility of using AI themselves to authenticate coins, the authorities have a particular interest in the study conducted at Concordia because the patterns observed could enable them to improve the coins they produce.

"[The authorities] especially want to know what precise characteristics are being falsified in order to make their coins more difficult to copy," Khazaee said.

In addition to their work at Concordia University's Centre for Studies in Pattern Recognition and Artificial Intelligence, the two researchers also head up a start-up company that uses AI in the field of sustainable development.

This report by The Canadian Press was first published in French on May 26, 2024.