Smart Chips Are Coming: With reports of the mass design and manufacture of chips that leave deliberate (or even undrilled) backdoors for intelligence services, large businesses and important individuals could be at risk from the use of a chip that, while seemingly harmless, can allow hackers to sabotage devices used in healthcare, public and financial infrastructure, military or government electronic devices.
Siddharth Garg, an assistant professor of electrical and computer engineering at the Tandon School of Engineering at New York University, and his fellow researchers are developing a unique solution: a chip that contains an embedded module that will prove that its calculations are correct and another external module that will validate the proofs of the first module.
So software viruses are easy to detect and fix with downloadable patches, if they are intentionally placed, are invisible and operate secretly. For example, a hidden “backdoor” function could allow hackers to modify or take over a device or system at a specific time. Garg’s system features a new approach called “verifiable computing” (VC). The new function keeps tabs on a chip’s performance and can detect signs of a potential Trojan.
The ability to control the components we put into our computers has become crucial in an age of distrust. These days, when a company designs and wants to build new hardware, the production costs are now so high that the designs are sent to offshore manufacturing companies, and security cannot always be guaranteed.
The system proposed by Garg and his colleagues features a verification processor that can be built separately from the chip. “It uses an external verification module from a trusted manufacturer, which means that even if the rest of the chip is made by an untrusted manufacturer, there will be a module that can present proofs of correctness,” Garg says.
So the company that designs the chips can turn to a trusted manufacturer to build a separate, less complex unit: an ASIC (application-specific integrated circuit), whose sole job will be to validate the proofs of correctness generated by the internal module of an untrusted chip.
Garg said that this arrangement provides a safety net to the end user's chip.
“Under the current system, I can get a chip that has a Trojan embedded in it. It might not show up during the first test,” Garg said. “But later on it could start misbehaving.”
