Engineer shows how to crack a 'secure' TPM chip
Demonstration at Black Hat describes difficult, but successful, attempt on Infineon chips
- By William Jackson
- Feb 02, 2010
A security engineer who reverse-engineered the family of chips from Infineon Technologies AG that includes its Trusted Platform Module implementation showed an audience at the Black Hat Federal Briefings how he cracked the chip and accessed its data.
Using an electron microscope to operate at the nanometer scale and Adobe Photoshop to plan his attack, Christopher Tarnovsky was able to sit on the chip’s data bus and “listen” to unencrypted code.
“This takes you somewhere that Infineon says you can’t go,” said Tarnovsky, who runs Flylogic Engineering and specializes in analyzing semiconductor security. He demonstrated his technique Feb. 2 at the conference, in Washington, D.C.
Invasion of the botnets: Cyberattacks on the rise
New threats arise from once-trusted protocols and services
Tarnovsky said he has accessed almost all of the code from a TPM chip and also had dumped data from secure licensing chips used in products such as Microsoft’s Xbox 360 video game console.
“Nothing inside that device is secure,” he said. “I can access all the keys and secrets on the chips.” He said he was not attacking Infineon specifically and that chips from other manufacturers might also be vulnerable to physical attacks. “These chips are not as secure as the vendors tell you they are.”
Tarnovsky said he had reported his successful attack to Infineon about a month ago but had not heard back from the company. Infineon did not immediately respond to a GCN request for comment.
But Tarnovsky said, “Infineon is not happy at all. They know I have broken the device.”
The device is the SLE 66PE family of contactless interface microcontrollers. In addition to securing commercial products that require licensed use, they also are used to implement the Trusted Platform Module, a set of specifications from the Trusted Computing Group for implementing cryptography in silicon. The chips can be used to support data protection, communications security, strong authentication, identity management, network access control and nonrepudiation.
The attack is not for the novice. It requires physical access to the chip itself as well as access to a Focused Ion Beam workstation, a type of electron microscope that can see at a much smaller scale than an optical microscope and can manipulate tiny needles less than a micron across, injecting conductors and insulators to rearrange the chips’ circuits.
“Don’t think that this is easy,” Tarnovsky said. He spent six months on the project and still has unanswered questions about the chips’ operations and security. The process of reverse-engineering would cost about $200,000 commercially, but he says that now that he has the technique worked out he can access a chip’s core and its data in six or seven hours.
“I can get any piece of information stored on the chip,” he said.
He began by buying chips in bulk for pennies apiece to experiment with and break. He stripped each layer off the chip to expose its topography, imaged the layers using optical and electron microscopes, and used Photoshop to layer the images so he could plan his attack through an intact chip.
His target was the processing core, the chip’s “central nervous system.” Instructions in the core have to run in the clear; information in the rest of the chip is encrypted.
Tarnovsky said the security in the chip, which is both physical and logical, is good.
“I really like them a lot,” he said of Infineon. “The security is built in layers. It’s like Fort Knox in there,” with defenses such as optical sensors that can detect light from optical telescopes.
The first layer of defense is a microscopic wire mesh covering the chip. Breaking the electrical circuits in the mesh would disable the chip, making it useless. Tarnovsky mapped the circuits and was able to plot a way to bridge them so that he could punch holes 3 microns across into the lower strata.
Once inside the core, “I can sit on the data bus and listen,” he said. “I can get any piece of information stored on the chip.”
Because of its complexity such an attack is not likely to become common soon. But not only is the data on individual chips at risk from a determined attacker, but once the manufacturer’s code is copied from the chip it could be used to produce counterfeit chips, which also could contain backdoors.
William Jackson is freelance writer and the author of the CyberEye blog.