How AI Can Help Stop Counterfeiters (Or Make Them More Effective Than Ever)

AI is proving useful in many applications ranging from games to medical diagnostics. However, there is a growing realization of the importance of AI in a field that can potentially save billions for business globally: counterfeits.

Many companies in the west face a real crisis with eastern countries such as China and Taiwan: counterfeit products. Companies invest millions in engineers, product invention, development, and even legislation in the hope to produce a product that will generate profit. It is this drive for profit that encourages technological advancements and therefore helps to push society into the future.

However, manufacturers in other countries including China will often ignore patents and copyright laws and produce counterfeits of popular products. Western companies, in particular, insist that this undercuts the patent holders and therefore has a negative impact on the profits of companies who invest in innovation. This negative impact ripples throughout the economy and results in the discouragement of new inventors to invest in their designs.

But counterfeits are not only bad for the designers of the original product. They're also potentially bad for the environment and health as many of these counterfeits do not go through any safety checks or certification for conformity to regulations such as CE and FCC.

Counterfeits are not limited to handbags and designer shoes, just about every industry is affected by counterfeits including the electronics industry. Counterfeit components can cause a headache for most (if not all) companies who are involved with sourcing parts. In fact, it was only last year that I published a news article on how to spot counterfeit ICs and the reasons why they should be avoided.

So, as a result of counterfeit products, the global economy severely suffers as well as consumers of counterfeit products.

Spotting counterfeits can be done by professionals and individuals alike using multiple techniques including image comparison, microscopic inspection, secret markings (such as invisible ink), and even authentic marks and damage (such as a football jersey that may have grass stains and tears). However, counterfeiters are becoming better thanks to modern manufacturing techniques and so traditional methods are no longer guaranteed to differentiate fakes from the real article.

An emerging technology, however, is stepping up to the task to fight counterfeit products: AI.

AI as a Tool Against Counterfeiting

AI systems are very good at reading databases and making many comparisons simultaneously. This makes them highly suitable for identifying counterfeit products which, in a general sense, involves comparing information from "genuine" items to those of possible counterfeits.

 

Feeding Visual Data to an AI Algorithm

Entrupy is a company that sells a microscopic camera system that is placed on the suspected counterfeit product and passes captured images to an AI algorithm. This algorithm makes checks against a database of authentic products to conclude if the suspected item is authentic or not. With an accuracy of 96.4% (and this figure is steadily rising), many customers (both big and small), have seen the benefit of such AI-driven systems and the company now has (as of September 2016) 130 paid customers.

While such anti-counterfeit systems work well, they are mainly aimed at fashion ware such as handbags and so electronic industries do not benefit from them. However, the concept behind Entrupy could possibly be migrated over to the electronics industry because companies are installing cameras into pick-and-place machines which scan a component before placing it down.

 

The Entrupy system can identify counterfeit designer bags with 97% accuracy. Image courtesy of Entrupy

 

If a component is determined to be counterfeit during device assembly, the machine could automatically inform the operators, managers, and those involved with the supply chain as to prevent potentially faulty devices from making their way into consumer hands.

 

GPU-Accelerated Computing

Another example of advanced counterfeit identification uses GPU-accelerated computing. One company, Cypheme, has used such techniques to identify counterfeit products. However, their origin story is more serious than most with the driving force not being money, but human life.

The mother of one of the four co-founders of Cypheme nearly died as a result of counterfeit medicine, spurring their mission against the counterfeit industry and the lack of manufacturer accountability for dangerous products.

Cypheme uses a rather interesting method for counterfeit detection that relies on co-operation with manufacturers of authentic products. The manufacturers' part is to integrate a traceable paper with a specific design on the product packaging. Then, any smartphone with the Cypheme app can be used to take pictures of the packaging and the camera will identify individual grains on the traceable paper. At this point, the grain information is then sent to an AI algorithm, powered by NVIDIA GPUs, to determine the authenticity as well as the date of manufacture, all in seconds.

The advantage of Cypheme is the lack of RFID or NFC ICs needed on the product packaging and the only requirement of the phone is the availability of an 8-megapixel camera.

The Flip Side: AI as a Counterfeiting Weapon

While AI has proven to be incredibly useful for fighting counterfeit products, it also brings about a serious problem because it can be reversed to produce almost perfect counterfeits.

Counterfeits are usually identifiable thanks to human error and lack of care. However, AI systems are meticulous by design when it comes to detail and may be able to produce counterfeits so good that they are indistinguishable from the real article.

 

AI that Can "Paint" like a Master

One case to support such a claim is the result of a partnership of Microsoft, ING, Delft University of Technology, and the Dutch museums Mauritshuis and Rembranthuis. Using AI and a 3D printer, the team was able to create a painting with brush marks, colors, and a subject that all resembled an authentic Rembrandt painting.

The AI algorithm used 346 paintings (which were also 3D scanned) to create a painting that represented the common themes of paintings of the Rembrandt era: a white male sporting a period-appropriate hat and moustache. The figure shown in the painting is not a copy or an alteration of someone from the past—it was wholly created by the AI.

 

The counterfeit painting produced by AI. Image courtesy of Microsoft/ING/Delft University of Technology

 

This experiment, while fascinating, also means that paintings "discovered' in the future must be treated with extreme skepticism. 

 

Counterfeiting Video Footage 

A second, more frightening case of AI, is a program called Face2Face which is in development by the Max Planck Institute for Informatics, the University of Erlangen-Nuremberg, and Stanford University. This application, utilizing deep learning, can generate real-time facial expressions from a user that are superimposed over a target video source.

What makes this software unbelievable is the accuracy and speed at which it can manipulate a video source to make it appear that a person is saying something. This may lead to cyber criminals and others to take video footage of individuals and put words into their mouths. Essentially, Face2Face allows for counterfeit video evidence.

Read More

• AI Predicted to Commit More Cyber Crime than People by 2040
• The Future of AI and Predictive Policing
• MIT Gives Computers the Capability to Predict the Future with Deep Learning

AI has the potential to mitigate various counterfeit industries, from components to pharmaceuticals. It could, however, also be the ultimate tool for creating extraordinary counterfeits. As it develops over the next few years, it will find more applications on both sides of the law.