How AI Is Helping Wage War on Counterfeit Goods

Counterfeit goods are big business; $460 billion worth of fake merchandise was bought and sold in 2016, according to the International Trademark Association, with sneakers, luxury handbags, electronics, and sunglasses among the most popular items.

In the wake of rampant imitations, technologists are developing tools that can help determine whether products are real or fake. Entrupy, a startup that uses a microscopic camera connected to a smartphone to take photos of the product for analyzation, recently added a mechanism that uses machine-learning algorithms to detect whether a product is real or not.

Goods made of materials like leather are easy to for its machine-learning algorithm to process, according to Entrupy CTO Ashlesh Sharma. "Reflective surfaces (such as glass) are difficult to identify since the technology is based on reflective optics," Sharma says. "But we are working on computer vision-based methods to lower the inaccuracies due to reflection."

Since the method is based on machine learning, AI won't be able to identify the newest items simply because they haven't been classified within its system. But Entrupy is hoping it can work with companies like Apple whenever it launches a new product.

"Counterfeits flood the market as soon as something of high demand is released, such as iPhones or high-end sneakers," says Entrupy CEO Vidyuth Srinivasan. "We plan to support some of these products as soon as they are made available or in certain cases, with cooperation from brand-owners."

Srinivasan's vision for Entrupy is to help facilitate buyer confidence during the purchase of physical goods. "We want to be part of any transaction where there is a trust deficit, especially in the online world, where people cannot touch/feel the product," says Srinivasan.

Other, larger entities, such as Alibaba, are utilizing big data techniques to identify fakes and stop them at their source, ameliorating supply chain issues. The company's software scans roughly 10 million products a day, though some argue that because counterfeiting is such a huge part of the Chinese economy—the country makes almost $400 billion annually from counterfeit goods—it's not likely to stop any time soon, and machine algorithms will need to keep up.

No current technologies are fail-safe, but Brian Markwalter, SVP of research at the Consumer Technology Association, says there will be a time when companies can expect technology to take AI and machine learning techniques to deeper levels, especially as smartphones become more powerful.

"There's lots of new work within the physics and chemistry sectors to develop new sensors," Markwalter says. "I'm seeing less innovation right now with machine learning. There are several companies, like Entrupy, that are using visible light to detect fakes. But after some more work in the field we can expect chemical sensors to help speed up that process. I've seen a couple of companies that are working on different types of scans. Coupled with better cloud-based computing, we'll have really powerful counterfeit detectors before long."

Despite these advances, Markwalter doesn't think the role of counterfeit experts will be diminished any time soon. "Humans have to guide systems," he says. "An expert might pick up an item and look at the material with a magnifying glass—which seems to be the application Entrupy is pursuing. But there's always an arms' race between the counterfeiter and the brand, and you need experts to sort through that."

This article originally appeared on