Hey there, human — the robots need you! Vote for IEEE’s Robots Guide in the Webby Awards.

Close bar

How to Catch a Memory Copycat

In the competitive world of memory chip manufacture, how do you find a fraud?

3 min read
How to Catch a Memory Copycat
Illustration: Randi Klett

In 2008, allegedly, a technician left SanDisk with a particularly good gift for his new employer—proprietary details about memory chips made by SanDisk and its partner on the project, Toshiba. Last Monday Toshiba revealed that it was suing the alleged receiver of that gift, SK Hynix, for US $1.1 billion and demanding that the company remove any chips from the market that use the trade secrets.

The chips in question are NAND flash memory chips, the nonvolatile memory of smartphones, tablets, USB drives, you name it. In this case the lawsuit is clear-cut: An employee allegedly downloaded and passed on files, and if the companies can prove it, the case is closed. But many times stolen trade secrets or patent infringements have to be found the old fashioned way—by reverse engineering.

“Just because somebody comes out with a competitive chip to your chip doesn’t mean anything’s been stolen,” says Terry Ludlow, the founder and CEO of Chipworks, a reverse engineering firm in Ottawa. “The first question you have to answer is: What have we got; where’s the evidence?”

“Given enough time and effort and money, you can pretty well figure out the whole thing.”

Chipworks helps companies uncover that evidence to show an “unreasonable likeness” between two chips, systems, or pieces of software, says Ludlow.

The first step is to get the chips in question—generally by purchasing them off the shelf. The reverse-engineers then take high-magnification photographs of the chip’s surface with a scanning electron microscope.

Getting a full picture of the compact and complex chips requires considerable patience. “Typically it depends on the level of technology and the degree of integration,” says Ludlow, “but these days there are usually a lot of layers of metal on top of the chip. You have to go layer by layer, scrape the layers of metal off the top for four, six, eight layers of metal.”

They take pictures all the while, and eventually form a photo-mosaic that recreates the layout of each layer of the chip, all pinned together in three dimensions.

“Given enough time and effort and money,” says Ludlow, “you can pretty well figure out the whole thing.” Of course, that’s rarely necessary. Normally the aim is just compare top levels of the two chips in question.

Jim Handy, a semiconductor market analyst at Objective Analysis, says that cases of outright trade secret theft are uncommon, but they happen: The one that stands out in his mind was from the late 1990s. (Evidently Samsung lost an entire design database to Taiwanese DRAM manufacturers, and the engineers who sold the information were overheard by a cab driver discussing the theft.) But according to Ludlow, cases of suspicious similarity are much more frequent, which is why reverse engineering stays in demand.

“The most interesting part of the whole trade secret thing comes back to: what’s mine and what’s my company’s?” says Ludlow. “There are people that test those limits constantly.”

The memory business is incredibly competitive. Handy says prices decrease at an average rate of 30 percent per year—and that means any loss of competitive edge can be hard to recover from.

“You have to have the most advanced technology in order to get your costs down to the point where you’re profitable at all,” he says. “And that’s hard to master. Memory companies struggle an awful lot with that—for it to slip through their fingers is a pretty big deal for them. That’s the reason why they’re seeking such large damages.”

“The most interesting part of the whole trade secret thing comes back to: what’s mine and what’s my company’s?” says Ludlow. “There are people that test those limits constantly.”

The Conversation (0)