Alitheon: Tackling food and label fraud with patented AI technology
17 May 2024 | Alitheon
Food and beverage fraud is on the rise and is reportedly costing US$40 billion annually via counterfeits, dilutions, substitution and mislabeling. Roei Ganzarski, CEO of Alitheon, a US-based software company, has developed patented AI technology to digitally ‘fingerprint’ goods and combat label fraud. He delves into the food labeling space and how AI can be used to tackle food fraud.
Welcome everyone, this is Liz Green.
Now food and beverage fraud is on the rise and is reportedly costing $40 billion US dollars annually via counterfeits, dilutions, substitution, and mislabeling.
Roy Ganzarsky, CEO of Alithon, a software company based in the US, has developed patented AI technology to digitally fingerprint goods and combat label fraud.
Roy joins us today to talk about food labeling and how AI can be used to tackle it.
Hi Roy, welcome, thanks for joining us today.
Thanks for having me.
Of course.
So what can you firstly tell us about how the company came about and why you decided to venture into the food space with your technology?
We started back in early 2017 to solve 4 big problems we were seeing in industry, and I use industry at large.
One was lack of a good system for traceability.
Where are my things?
Where do they go in the supply chain, in the marketplace, circular economy?
Are they in fact coming back, etc.
Second problem was a growing problem of counterfeits, bad people making things look like the real items in order to cheat and make money illegally.
Third problem was gray market items, even a A more difficult problem because gray market items are real items being sold illegally.
Think a pharmaceutical that's expired.
Someone changes the date and sells it as if it's new.
And then 4th, last problem was human error based on misidentification.
I have a bunch of real legal items in front of me.
They all look the same, but maybe they have different functions, different codings, different purposes, and I by mistake take the wrong one.
So those are 4 problems we wanted to solve, and we wanted to solve it in a way that wasn't the way.
It was being done today because the way it's being today doesn't work, which is why in food and beverage, for example, it's such a high dollar amount and high level of fraud, same as other industries, we didn't want to rely on additives or proxies, meaning stickers, labels, holograms, tags, ink.
All those things can be manipulated, changed, redone, removed, and even counterfeited themselves.
So relying on a sticker.
To tell you if something is real at best tells you the sticker is real, and at worst it's completely faked altogether.
So that's why we came to be, and food and beverage is really a natural area to enter into because it's something that we all put into our bodies.
It's not a purse, a watch, you know, a car part.
Which all of which, by the way, we do as , but these are things that we put just like pharmaceuticals and medical equipment, we put into our bodies.
We ingest them and they will have an impact.
And if they're good, great, but if they're bad, that could be a huge risk to health and life.
And so we wanted to make sure that we participate in getting rid of that problem.
Mhm, OK, really fascinating.
So why then are traditional packaging identifiers such as expiration dates and batch numbers so often targeted by fraudsters, do you think?
It's really easy.
If you think of things that are laid on the package, and it doesn't matter if it's a a bottle of beer or wine or a package, a bag of chips, it doesn't really matter.
Anything that's added to that package or to that label can easily be changed, be it ink that was used to print a date or a serial number.
I can take some alcohol rub, erase it, and print a new one.
Even easier if it's on a sticker.
Just peel off that sticker and put on a new sticker that you printed yourself, including, by the way, holograms.
You can go online onto eBay or Google or any one of those and do a search for hologram stickers or how to make your own, and you'll get a bunch of stickers that look like the ones you might find on a package, and you could just create your own.
And so the reason they're targeted is it's much easier to manipulate and fake a printed ID or a label than it is the actual product.
And people are gullible.
If we see a printed number, a date that looks right, a nice shiny hologram, we want to believe that things are what they say they are, and we'll buy them and use them.
And so what these bad actors are doing are taking advantage of our trusting gullibility by changing a few items that are printed or stuck on and and convincing us that whatever we're buying and putting into our bodies is good, when in fact it may be very bad.
Mhm.
Absolutely.
OK, really interesting.
Now AI and technology in general, we see is coming on leaps and bounds.
Why do you think now is the right time for Elitheon to develop its patented technology?
So there's a few things that happened.
One is as a company, our goal was to enable people to be able to do what we allow them to do in a simple fashion, meaning standard off the shelf cameras, no special expensive cameras and lighting that used to be required to do things kind of similarly, but rather we want to be able to do it with a cell phone, with a mobile phone.
So we had to get to a point in our world where mobile phone cameras were 1, good enough, but 2, accessible enough and cheap enough.
We are at that point.
Second was the ability to develop our algorithms in a way that allows us to do what we wanted to do, which meant no need for knowing what the item is.
So we don't have to train the kind of the proverbial word used to talk about machine learning or AI.
We didn't want to have to train the system on types of items because there are so many types of items, even within food and beverage.
So if you have to train the system on every one of them, that becomes problematic.
So we had to develop algorithms to resolve that, and we also wanted to make sure that when we go into production, there are zero false positives.
Now, what do by that?
If I were to tell you a real bottle of wine is fake, and I'm wrong, that's a false negative.
You've lost some money.
If I were to tell you that a fake bottle of wine is real and I'm wrong, that's a false positive, that could be detrimental to your health because you are drinking that bottle, right?
And so we wanted to make sure that when we go into production in any industry, let alone food and beverage where you're putting things in your body, there are zero false positives.
And so The time came together where you could see this confluence, this meeting of the algorithms that we were able to develop, together with low cost, affordable, really good quality cameras on mobile phones.
You bring those two things together and suddenly you're at a point where a litheon can do what we do.
Mhm, OK, fantastic.
We've kind of touched on it a little bit, but how can it be, how can this kind of technology be used to support food ingredient and packaging companies in the future then, because it's obviously quite, it's quite a niche, space, I would, I would say.
A lot of food and beverage is being sold worldwide on a regular basis with, as you said in your opening, a lot is being counterfeited.
If you think of just the dollar value, that's one thing.
Think of the amount of food and beverage that means when you look at that dollar value.
That means a lot of people are ingesting bad things.
So the next time you eat something and you get sick, you or worse.
Maybe it's not just some bug that got in or some virus that you caught.
Maybe it's bad food or bad drink that was sent there as counterfeit or gray market because it expired and you were duped into using that.
And so the way this can be used to solve these problems is really about traceability and transparency.
When you as a consumer, or I as a consumer, go and buy a can of beer, a bag of chips, a bag of a box of frozen food, it doesn't matter what.
I want to be able to believe that whatever is on that box that said or that bottle, and that's what it says, where it came from, when it was made, what the expiration date, I want to be able to know that what's printed or listed on that package is in fact true.
The challenge is, as we said before, it's so easy to manipulate that.
That as a consumer, I may not know.
Now these bad actors are taking advantage of people's, I'll call it naivete, gullibility, automatic trusting.
They take advantage of the fact that when we see a nice package with a nice hologram sticker on it and nice numbers printed on it and a date that's valid, we like to believe and we want to believe that it's good because why would anyone cheat us with food.
People cheat with food and beverage because it's easy and a lot of money could be made.
So, AI or what we do now in terms of vision AI will enable a consumer to pull out their cell phone, take a picture of that package.
And the data itself that came from the source, the origin of the manufacturer, will be able to say, this is in fact the package that I put together in this factory.
I'll even be able to look at a map and see where that was from this factory on this date.
This was the expiration date that the manufacturer gave, not that someone printed on a box, but here's the the expiration date that the manufacturer gave, and I can look at that on my cell phone.
And if it doesn't match what's on the box, I know not to buy it.
Now imagine that happens enough to where now these bad actors say, it's not worth trying to counterfeit or gray market that product because it's feature printed.
There's no way we can get away with it, so let's not do it altogether.
So while we may not be able to eliminate fraud within the food and beverage industry, it should definitely reduce it because anything that is feature printed, i.e., digital IDs is reported from it.
You won't be able to cheat the system, and that's where we hope to be going to.
Mhm.
OK, fantastic.
Yeah, there's definitely a lot of positives there.
What about any challenges or limitations that this might have?
Yeah, so, we can't, for example, because we are based on the technology of vision AI, meaning we take a picture of something and our algorithms from that image are able to extract the digital fingerprint or what we call a feature print.
Imaging is important.
I can't take a picture, for example, of a liquid, or a powder, or a gas.
And so, the best I can do, for example, on a bottle of wine, is make sure that the bottle is what we think it is, that the cork is what we think it is, and in fact that they're linked together.
That cork was put in that factory onto this bottle and has never been changed.
That we can do, but what we can't do is say the contents of that bottle are what they should be, because we can't take a consistent picture of liquid without it changing all the time.
So we can't do liquid powder or gas.
But the packages that these things will come in are able to be protected in such a way that the likelihood of someone being able to cheat the system becomes very, very low.
So that's one limitation or one aspect.
The second is we can't do diamonds and gemstones, which are not relevant to food and packaging because most people don't eat diamonds or gemstones, but the reason we can't do them is diamonds and gemstones are cut to be really sparkly.
And they are sparkly, meaning because they refract light really.
So if I took 3 sequential pictures of a diamond, for example, and wanted to look at them, they will all look completely different because the minute changes in light will completely refract differently on the diamond.
It makes it difficult with a phone.
You could do it with specialized cameras, but again, as I said, as a company, we want to be able to do this with phones so that it's accessible to everyone.
So those are the only limitations to our technology.
Other than that, We could pretty much do everything, paper, plastic, wood, metal, steel, gold, 3D printed items, and in fact live items as , although we tend to stay away from that, but other than those items that I mentioned, everything else pretty much goes.
OK, the sky's the limit then, fantastic.
And what about then the importance of providing track and trace products through a means to track and trace even, through supply chains.
How important is that and how.
How have you moved in in that space?
Yeah, it's, it's really critical.
Like if you think back to, to what I mentioned in terms of traceability, being able to know that what I'm picking up off the shelf or out of the fridge in the supermarket is in fact what it should be and what I think it is based on what's written on it.
The only way for me to really know that's true is through traceability, which means that from the point of origin.
It was actually somehow recorded, documented, and put in play, and every milestone on the way from the factory to the ship, from the ship to the truck, from the truck to the bicycle, from the bicycle to the shelf, it's all been.
Recorded a track so that when I pick up that box and take a picture of it, I can in fact know, yes, this is the item that went through that entire path, not because it's printed that it did, because again, anyone can print anything, but because I can know from the original manufacturer that that is what it is.
