A team of security researchers from the U.S. and Europe has released a paper showing how integrated circuits used in computers, military equipment and other critical systems can be maliciously compromised during the manufacturing process through virtually undetectable changes at the transistor level.
As proof of the effectiveness of the approach, the paper describes how the method could be used to modify and weaken the hardware random number generator on Intel's Ivy Bridge processors and the encryption protections on a smartcard without anyone detecting the changes.
The research paper is important because it is the first to describe how someone can insert a hardware trojan into a microchip without any additional circuitry, transistors or other logic resources, said Christof Paar, chairman for embedded security, Department of Electrical Engineering and Information Technology at Ruhr University in Germany.
Hardware trojans have been the subject of considerable research since at least 2005 when the U.S. Department of Defense publicly expressed concerns over the military's reliance on integrated circuits manufactured abroad, Paar said.
Often, the individual circuit blocks in a single microchip are designed by different parties, manufactured by an offshore foundry, packaged by a separate company and distributed by yet another vendor. This kind of outsourcing and globalization of chip manufacturing has led to trust and security issues, the paper noted.
Over the years, more attention has been paid on finding ways to detect and defeat hardware trojans deliberately introduced during the manufacturing process, especially in the case of chips used for military and other critical applications.
Somewhat surprisingly, less attention has been paid to how someone might build and implement such hardware trojans in the first place, he said.
Previous research papers have described hardware trojans consisting of small to medium-sized integrated circuits added to a chip during what is known as the hardware description language layer of the manufacturing process.
In contrast, the latest research shows how a hardware trojan can be introduced at a later stage of the design process by changing the "doping" on a few transistors on the chip.
Doping is a process for modifying the electrical properties of silicon by introducing tiny impurities like phosphorous, boron and gallium, into the crystal. By switching the doping on a few transistors, parts of the integrated circuit no longer work as they should. Because the changes happen at the atomic level, "the stuff is hard to detect," Paatr said. "If you look at it optically there is nothing different," so the trojan is resistant to most detection techniques.
Security researcher and cryptographer Bruce Schneier on Monday called the sabotage the researchers describe "undetectable by function testing and optical inspection."
The most "devastating" use of the technique is to modify a chip's random number generator, Schneier noted in a blog post. "This technique could, for example, reduce the amount of entropy in Intel's hardware random number generator from 128 bits to 32 bits," Schneier said.
"This could be done without triggering any of the built-in self-tests, without disabling any of the built-in self-tests, and without failing any randomness tests."
So while users assume that the random number generator is producing strong 128-bit encryption keys, in reality, it is generating 32-bit keys that can be easily broken, Parr noted.
There are several other scenarios where an integrated circuit can be modified to make it function in an unexpected fashion, he said. Detecting the modifications would require an additional level of testing of circuits, he added.
This article, Security researchers create undetectable hardware trojans, was originally published at Computerworld.com.
Jaikumar Vijayan covers data security and privacy issues, financial services security and e-voting for Computerworld. Follow Jaikumar on Twitter at @jaivijayan or subscribe to Jaikumar's RSS feed . His e-mail address is email@example.com.