Semiconductor company Microsemi has issued a statement denying that one of its products, a popular silicon chip called ProASIC3, has a backdoor built into it.
The ProASIC3 field-programmable gate array (FPGA) — a chip designed to be configured and programmed by customers according to their needs — has no designed feature that would enable circumvention of the user security, Microsemi said Thursday in a message on its website.
ProASIC3 chips are integrated into systems used in many industries, including the military, for various applications. The chip is marketed by Microsemi as having one of the highest levels of design security on the market.
University of Cambridge Ph.D. candidate Sergei Skorobogatov and Christopher Woods, a hardware security researcher at U.K.-based research company Quo Vadis Labs (QVL), claimed that they discovered an undocumented function in the ProASIC3 FPGA that can be used by an attacker with physical access to the chip to extract the intellectual property (IP) stored on it, despite such information being encrypted with a user-defined 128-bit AES key.
Skorobogatov and Woods were planning to release in September at the Cryptographic Hardware and Embedded System workshop a research paper that explains how they found the backdoor. However, a draft version of their paper leaked online and was used as a source for news stories earlier this week.
In the draft version, the researchers explain that they used a technique called Pipeline Emission Analysis (PEA), patented by QVL, to significantly increase the efficiency of differential power analysis (DPA) methods.
DPA attacks can be used to extract cryptographic keys from hardware devices by analyzing fluctuations in their power consumption during normal operation. However, traditional DPA methods are not very accurate and usually require a lot of time and expensive equipment to use successfully.
Using the new PEA technique, the AES key can be extracted from ProASCI3 chips in seconds instead of hours, and a separate user-defined passcode that protects their configuration settings can be extracted in hours instead of years, the researchers said.
In addition to successfully extracting the chip’s keys, the researchers claim to have also identified a function in the device’s memory that can be used to unlock and reprogram virtually any information from the chip.
Access to this function was protected with a third key, which was successfully extracted in about a day by using the PEA technique, the researchers said.
The function discovered by the researchers is a privileged internal test facility reserved for initial factory testing and failure analysis, but it is disabled in all shipped devices, Microsemi said on Thursday.
Furthermore, the function can only be accessed on a customer-programmed device only if that customer’s passcode is also supplied, the company said.
“Customers have an option to program their chosen passcode to increase the security; however, Actel/Microsemi does not tell its customers that a special fuse must be programmed in order to get the backdoor protected with both the passcode and backdoor keys,” the researchers said in a statement on Friday.
Also, even if a customer passcode is used to protect the backdoor function, that passcode can be recovered in hours with the PEA technique, the researchers said.
Microsemi could not confirm if the key extraction attack described by the two researchers is possible, because it didn’t have access to the technology and hardware setup they claim to have used, the company said.
In response, the researchers pointed out that the PEA technique is described in QVL’s patent and is, therefore, publicly available.
Microsemi advises customers who are concerned about DPA-type attacks to program their devices with the highest security setting. “This security setting will disable the use of any type of passcode to gain access to all device configurations, including the internal test facility.”
However, the researchers claim that this is achieved through a mechanism called Permanent Lock, which is vulnerable to fault attacks and can also be disabled by an attacker.