How chip-level chaos can help secure devices
Researchers at Ohio State University are introducing chaos into computer chips to develop digital fingerprints that may be unique enough to foil even the most sophisticated hackers.
The new solution makes use of an emerging technology called physically unclonable functions, or PUFs, to take advantage of tiny manufacturing variations found in each computer chip. These slight variations – sometimes seen only at the atomic level – can create unique sequences of 0s and 1s that PUF researchers in the field call “secrets.”
Currently PUFs have a great but limited number of secrets, making them vulnerable to hackers that have the time and technology to break them. However, the Ohio State team has “found a way to produce an uncountably large number of secrets to use that will make it next to impossible for hackers to figure them out, even if they had direct access to the computer chip,” Daniel Gauthier, senior author of the study and physics professor, told Ohio State News.
With funding from the Army and the Ohio Federal Research Network, the researchers created a complex network in their PUFs using a web of randomly interconnected logic gates, which normally take two electric signals and use them to create a new signal.
“We are using the gates in a non-standard way that creates unreliable behavior,” Gauthier said. “But that’s what we want. We are exploiting that unreliable behavior to create a type of deterministic chaos.”
The chaos amplifies the chips’ manufacturing variations, allowing the team to change the number of secrets that are being produced, according to Noeloikeau Charlot, lead author of the study and a doctoral student in physics at Ohio State.
“Chaos really expands the number of secrets that are available on a chip. This will likely confuse any attempts at predicting the secrets,” he said.
The researchers calculated that their PUF could create 1077 secrets – so many that it would take a hacker longer than the life of the universe to guess every secret available in that microchip, Gauthier said.
The researchers have demonstrated that their PUF can resist machine learning attacks, including deep learning-based methods and model-based attacks, and they are now asking other research groups to see if they can hack it.
These new PUFs could potentially be used to create secure ID cards, to track goods in supply chains and as part of authentication applications, where it is vital to know that you’re not communicating with an impostor, Gauthier said.
“It is a constant battle to come up with technology that can stay ahead of hackers,” he said. “We are trying to come up with technology that no hacker – no matter your resources, no matter what supercomputer you use – will be able to crack.”
The study was recently published online in the journal IEEE Access.
Connect with the GCN staff on Twitter @GCNtech.