Chip Puzzle: DARPA will order an unbreakable processor

Earlier this year, information on processor vulnerabilities in Meltdown and Specter became the first high-profile cybersecurity topic in 2018. The insecure cache of software commands for Intel, AMD, and Arm chips issued over the past 20 years has led to a galaxy of lawsuits and is still being discussed in the media. According to the Department of Defense Advanced Research Projects (DARPA), 40% of all software exploits can be avoided by eliminating hardware flaws.

Therefore, the Office was thinking about developing an unbreakable processor that would make exploits impossible. The grant for the creation of the chip was received by a team from the University of Michigan. In the article we will analyze what is already known about the processor.

/ Flickr / jin / cc

How does the chip work?

The MORPHEUS project is one of the initiatives implemented under DARPA grants under the SSITH program , which aims to eliminate hardware vulnerabilities. The concept of an uncrackable processor is based on an approach that involves not responding to known attacks (“ patch and pray ”), but building a universal system that can cope with any threats.

Todd Austin, a computer science professor at MORPHEUS, compared the hacking of a future chip with an attempt to collect a Rubik's cube, whose edges rotate each time a person blinks.

In order for an attacker to gain access to the system, he first needs to find a software bug (what was the error in the program command cache). After that, it remains to determine where the valuable data is stored, and to "pick up" them. Due to the architecture of the MORPHEUS processor, the detection of the “location” of the vulnerability and data becomes impossible. Even if an attacker finds an error and tries to use it, the vulnerability and valuable information will “change their location” randomly. In case the attacker proves to be fast enough, a second level of protection is provided in the form of encryption and domain enforcement policies that will guarantee additional obstacles.

/ Pxhere / CC

Such an approach can protect both hardware and software, and solve the problem of “ zero-day vulnerability, ” that is, protect computers from future threats that have not yet arisen.

The ExtremeTech editor recalls the Rowhammer exploit when talking about the Rubik's Cube chip . A hacker could run a program that repeatedly accessed specific rows in memory, as if “tapping them with a hammer”, until electromagnetic radiation penetrated into a neighboring section of memory and changed the value of individual bits.

As a result of such an attack, an attacker could increase his access rights in the system. For example, in the image below, an attack can be fired on either the purple line to “flip” the yellow bits, or the yellow lines to “flip” the purple bits.

/ Wikimedia / Dsimic / CC
But for a chip that can change memory addresses and store data in encrypted form, this “selection method” is not terrible. According to Austin, security systems like MORPHEUS have not yet been implemented, since they are too expensive to do as software. Scientists from Michigan hope that DARPA support will allow them to create hardware-level protection that will not be costly. However, it is not yet clear what resources will be required to integrate the technology into modern processors.

Other developments

At the end of last year, it also became known about the creation in South Korea of ​​a similar solution to protect the system. It is based on the method of “ physically nonclonable function ” (PUF) - a function that is implemented in the physical structure and is easy to evaluate, but difficult to fake. The Korean development uses nanoelectromechanical systems consisting of a silicon nanowire, which is suspended in a liquid between two gates - one and zero. During manufacture of the PUF, the fluid in which the nanowire floats evaporates, and the wire randomly “sticks” to one of the gates. As a result, a security code is generated that cannot be picked up from the outside.

Another prototype system with “physical passwords” is being run in Abu Dhabi. According to Ozgur Sinanoglu, Deputy Dean for Engineering at New York University Abu Dhabi, their chip is a solution with a hacking system integrated into the hardware. Special technology blocks hardware and allows access to the chip capabilities only to those who know the secret key. Until then, the purpose of the chip remains unknown.

A few more materials about IB from the 1cloud blog: