The Center for Data Innovation spoke with Roei Ganzarski, CEO of Alitheon, a U.S.-based company focused on authenticating, identifying, and tracing physical items, parts, and products using just a digital photo. Ganzarski discussed the potential of using advanced optical AI to track and trace items in production lines and supply chains to stop counterfeits and other fakes.
Daniel Castro: How did Alitheon get started, and what is the company trying to achieve?
Roei Ganzarski: Alitheon was started late 2016 by experienced optical mathematicians to develop deep tech machine vision solutions that could help companies authenticate, identify, and trace individual physical items. When you don’t know if a part or product you are about to use are real, or authorized, or even the right one to use, that can cause tremendous financial, brand, and safety risks. This is a real problem in industry that is now costing $2.3 trillion—yes with a T—a year even though companies spend close to $140 billion a year trying to solve it. Today the world relies mostly on “additive” solutions like labels, stickers, etc. Alitheon was set up to create a solution that did not require adding, manipulating or even touching an item. A solution in which the item itself, is its own security. We did it!
Castro: What challenges did the company face to create technology that reliably identifies physical items using only a photo?
Ganzarski: As you might imagine, developing this kind of technology was very difficult. It had to identify individual items out of many identical items without false positives. And it had to do so without adding, marking, or touching the item. And the solution had to be simple and easy to use. Moreover, we didn’t want to rely on or even use machine learning since that meant having to “train” the system on each new type of items which would take away the simplicity of the system we wanted. We needed to be able to replace all the flaws of current labels and for that you need advanced optical AI—the ability to take images of an item and more importantly to interpret what is seen and captured in the image in such a way that it can create a unique type of identifier (an item fingerprint of sorts) and then identify the one of the millions (like a label does), but without the ability to be moved, changed, or faked. It must be scalable and repeatable and easy to use. These combinations are not simple to achieve. This meant working in a complex intersection of optical, software, and advanced mathematics and physics. These are not the type of problems solved in a month or even a year. But the team solved it and the system is pretty amazing.
Castro: What are some of the benefits of using photos to authenticate, identify, or trace products?
Ganzarski: Labels, barcodes, etc. on items have a few inherent flaws. They can be damaged, erased, fall off, scratched off, etc thus leaving you no way to identify them. They can be moved from one item to the other. Thus you see a real label but it is on a fake product or on someone else’s product and now you have nothing. They can be faked—fake label on a fake product.
Castro: How can different businesses in the supply chain use Alitheon’s technology to protect consumers from counterfeits?
Ganzarski: Counterfeits have a very one-sided interpretation but we look at the world of counterfeits in three lenses.
First, “intentional counterfeits”: a “bad actor” intentionally tries to mimic your product in order to sell it instead of yours. We are not talking cheap knockoffs, rather sophisticated counterfeits.
Second, “midnight runs”: a “bad actor” in or related to a supplier, intentionally makes more of your product and sells it in other market streams. These are real authentic products or items, but they are unauthorized and mostly illegal. Even the best of the machine learning based anti-counterfeit systems won’t work here since the item is in fact a “real” item, just not authorized.
Third, “mistaken identify”: an authentic part, no bad actor, but the part cannot be identified so the wrong part is used.
In all of these cases, tremendous financial, brand, reputation, and safety risks exists. Be it a counterfeit luxury watch, fake medicine, fake brake pad, the wrong item mistakenly used in a hospital or on an airplane, the problem and consequences are growing each year.
Alitheon’s FeaturePrint system focuses on identifying each individual item. Think of a human fingerprint—if I have your fingerprint, any time in the future, anywhere in the world, I can identify who you are and where you have been by that fingerprint. Not just identify that a “human” was onsite, but rather you specifically, and no one else. Imagine that level of specificity in your supply chain that you can irrefutably identify your items, products, parts and make sure the right part, the real part, is being made, sent, received, installed. All by just taking a picture with a phone or an off-the-shelf industrial camera. Not just knowing the shoes are real, but they are authorized and going to the right location. Not just that the medicine is real when it left the factory, but when it arrived in the pharmacy and right before it is given to a customer. Imagine being able to do so without the fear that the “safety sticker” or “sophisticated hologram,” or bar code are themselves fake.
Castro: In addition to securing the supply chain, what are some of the other potential applications of Alitheon’s technology?
Ganzarski: Alitheon’s FeaturePrint system is deployed in supply chains but also in production lines to track and trace items as they go through multi-step production or assembly lines; in secondary marketplaces to make sure that the used or collectible item you are buying is indeed what it is supposed to be; in retail to ensure high-end expensive items are indeed original from the manufacturer; in pharma to ensure the right products being made by the pharma, are what end up in your body and nothing else; and in automotive and aerospace manufacturing to make sure only authorized parts get installed on your car or aircraft.