Startups Alitheon and Cybord are tackling the problem of counterfeit chips in the electronics supply chain.

Unsurprisingly, the counterfeit problem mushroomed during the chip shortage caused by the Covid-19 pandemic. “When the shortage suddenly appeared, companies were willing to accept processors, chips, boards and other products that had the serial numbers and the box stickers literally ‘Sharpied’ out,” Alitheon CEO Roei Ganzarski told EE Times. “They were willing to accept them because they needed to deliver product.”

And, new shortages may soon emerge, added Cybord CEO Oshri Cohen: “China is heading now to new export restrictions related to gallium and germanium. If China is going to restrict this, we definitely are going to see restricted supply chains very soon. I don’t think that that we can find any substitutes because, again, the process of developing these materials is very long. You need at least 5 to 10 years to develop such new materials.”

The two startups are using different methods of “fingerprinting” to improve traceability and eliminate potentially disastrous system failures caused by the adoption of outdated or sub-standard components.

Cybord in 2019 spun off from a $1 billion project in the Israeli Defense Forces that was nearly scrapped because of a faulty capacitor. Cohen gave Texas Instruments (TI) as an example.

“Texas Instruments is something that counterfeiters prefer,” he said. “The easiest way to make a counterfeit is to take a Texas Instruments disqualified part and change the date… change the marking.”

TI doesn’t deny that counterfeiting is an issue.

“To help mitigate the risk of counterfeit products, TI strongly encourages its customers to purchase all TI semiconductor products through TI’s authorized sources: either directly from TI or through a TI-authorized distributor,” Ellen Fishpaw, a TI spokesperson, told EE Times. “TI is unable to guarantee the reliability and authenticity of parts purchased outside of authorized channels.”

Still, the issue is bigger than just one chipmaker, Ganzarski said. “Any major brand that you can think of has counterfeit and gray market problems.”

Counterfeit chips are a growing problem that requires more effort to “police,” Semiconductor Advisors President Robert Maire told EE Times.

Defective devices installed in critical equipment like automobiles can result in fatal failures, he noted.

“There’s really not a good audit trail of tracking a chip from design to manufacture to distributor to end customer,” he said.

Methodologies differ

Alitheon and Cybord use different methods to spot counterfeit chips.

Alitheon “fingerprints” manufactured items by taking a picture of each one, including its distinguishing characteristics, as it rolls off an assembly line.

“I can serialize, and just like with people, if I fingerprint items, I can say which item it is. I can say something is fake because clearly it isn’t fingerprinted,” Ganzarski said.

The company examines up to 5,000 points of interest on a product compared with the 11 points of interest used to identify a human fingerprint. “FeaturePrints” can be used to trace each item as it goes through a supply chain, he said. Rather than using an AI methodology to identify chips, the company employs discrete mathematics.

“Machine learning, by definition, can’t serialize items,” Ganzarski said. “It can only look at class level. This is a microprocessor, or it isn’t a microprocessor. It solves a counterfeit issue, but it doesn’t solve traceability. It doesn’t solve human error. It doesn’t solve the gray market.”

The system works with a range of products that can include processors, PCBs or servers, he said.

“We are product-agnostic by design. Whether it’s a microprocessor, a baseball card or a brake pad—you can use the same system.”

Alitheon investors include BMW, Ganzarski noted.

A variety of companies in different industries, such as automotive, precious metals, defense and fashion, are using the company’s tech.

The Fast Sneaks Project, a non-fungible token (NFT)-based sneaker maker founded by artist Brendan Murphy, is using Alitheon’s technology to authenticate its collectible shoes. Each pair is linked to an NFT, facilitating ease of trading between collectors.

Cybord deals only with chips when they are assembled on a PCB. The company uses an AI system focusing on as many as 60 parameters.

“We take an image of a component,” Cohen said. “We scan it pixel by pixel. We are using AI. We can tell the database, ‘This is how these components look when it’s a good component, and this is how it looks when it’s a bad component,’ and we train the models.”

Cybord customers include companies selling datacom equipment and solar panels, Cohen said.

“They are providing their customers 20 to 25 years of guarantees,” he said. “The products are installed over roofs, universities and so on. If they have something bad in the field, it costs them a fortune to bring it back and fix it.”

Early warning system far away

The counterfeit detectives need more time to develop an early warning system of global scope.

“It’s definitely early stage,” Cohen said. “We have more than 4 billion components in our database. We’re gathering an additional 250 million every month. Four billion components [are] not enough to show a global trend. [We need] to see maybe 400 billion before we can start indicating global trends.”

To counter counterfeiting, the ideal is for an original manufacturer to stop the problem at the source, Ganzarski said.

“If you want traceability or the ability to track or the ability to avoid gray markets, you can do it anywhere along the supply chain. Imagine even the simplest example: I’m a retailer. I sell you a board. I can, in fact, register the board that I’m selling. If you come back a week later and say, ‘Hey, I’m returning the board,’ I can check. Are you returning to me what I gave you?”

It’s still possible that the netlist that a chip designer sends to a foundry can be intercepted and modified so that a chip contains a hidden backdoor that hackers could exploit.

“There’s really no way of telling when you send this random sequence of numbers that the chip disables itself, blows up or whatever,” Maire said. “You could easily build in something like that with feasible links.”

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved