Research Bits: Aug. 8

Speeding NVM encryption

Researchers from North Carolina State University propose a way to speed up encryption and file system performance for non-volatile memory (NVM).

“NVMs are an emerging technology that allows rapid access to the data, and retains data even when a system crashes or loses power,” said Amro Awad, an assistant professor of electrical and computer engineering at North Carolina State University. “However, the features that give NVMs these attractive characteristics also make it difficult to encrypt files on NVM devices — which raises security concerns. We’ve developed a way to secure files on NVM devices without sacrificing the speed that makes NVMs attractive.”

Existing methods for file system encryption use software, which was not a problem as it didn’t slow data access times. “But now that NVMs are allowing faster access to file data, the software approach to file encryption has become a problem, because it slows down overall operations,” said Kazi Abu Zubair, a Ph.D. student at NC State. “To address this challenge, we’ve developed a novel architecture that incorporates some elements of the encryption and decryption process into hardware, which is faster than software. As a result, processes that allow users to store and retrieve file data securely are significantly faster.”

In simulations, the researchers found that using their novel encryption architecture to secure files in NVMs slowed down operations by 3.8%, when running workloads that were representative of real-world applications. When using software approaches to provide security for the same workloads, operations slowed by about 200%. “If this was implemented in commercial processors, it would significantly improve performance for secure file operation in large data centers and cloud systems,” Abu Zubair said.

“While this work addresses file encryption, we think it is important to assess other security functions — such as auditing and run-time ransomware detection –in the context of direct access file systems,” added Awad. “And addressing those security functions using traditional software approaches can also slow system performance. We’re optimistic that our hybrid hardware/software approach may be able to improve performance for those functions as well — that’s an area we’re exploring.”

Magneto-electric transistor

Researchers from the University of Nebraska Lincoln and University at Buffalo developed a magneto-electric transistor. They say their design approach is the first step in testing new combinations of magneto-electric and 2D materials, with the potential to reduce both power and area.

“The traditional integrated circuit is facing some serious problems,” said Peter Dowben, professor of physics and astronomy at Nebraska. “There is a limit to how much smaller it can get. We’re basically down to the range where we’re talking about 25 or fewer silicon atoms wide. And you generate heat with every device on an integrated circuit, so you can’t any longer carry away enough heat to make everything work, either.

“So you need something that you can shrink smaller, if possible. But above all, you need something that works differently than a silicon transistor, so that you can drop the power consumption, a lot.”

The team’s device relies on spin, a magnetism-related property of electrons that points up or down and can be read, like electric charge can, as a 1 or 0. The team turned to graphene, in which electrons through can maintain their initial spin orientations for relatively long distances. They combined it with the magneto-electric material chromium oxide, in which the spins of the atoms at its surface can be flipped from up to down, or vice versa, by applying a small amount of temporary voltage.

They explained, “When applying positive voltage, the spins of the underlying chromium oxide point up, ultimately forcing the spin orientation of the graphene’s electric current to veer left and yield a detectable signal in the process. Negative voltage instead flips the spins of the chromium oxide down, with the spin orientation of the graphene’s current flipping to the right and generating a signal clearly distinguishable from the other.”

“Now you are starting to get really good fidelity (in the signal), because if you’re sitting on one side of the device, and you’ve applied a voltage, then the current is going this way. You can say that’s ‘on,’” Dowben said. “But if it’s telling the current to go the other way, that’s clearly ‘off.’ This potentially gives you huge fidelity at very little energy cost. All you did was apply voltage, and it flipped.”

Dowben noted that there are many other alternatives to graphene that may be more suited to a magneto-electric transistor, and looks forward to other researchers trying other combinations. “Now that it works, the fun begins, because everybody’s going to have their own favorite 2D material, and they’re going to try it out,” Dowben said. “Some of them will work a lot, lot better, and some won’t. But now that you know it works, it’s worth investing in those other, more sophisticated materials that could. Now everybody can get into the game, figuring out how to make the transistor really good and competitive and, indeed, exceed silicon.”

Honey memristor

Engineers from Washington State University developed a memristor using honey.

“This is a very small device with a simple structure, but it has very similar functionalities to a human neuron,” said Feng Zhao, associate professor of WSU’s School of Engineering and Computer Science. “This means if we can integrate millions or billions of these honey memristors together, then they can be made into a neuromorphic system that functions much like a human brain.”

The memristors were created by processing honey into a solid form and sandwiching it between two metal electrodes, making a structure similar to a human synapse. The researchers tested the honey memristors’ ability to mimic the work of synapses with high switching on and off speeds of 100 and 500 nanoseconds respectively. The memristors also emulated the synapse functions known as spike-timing dependent plasticity and spike-rate dependent plasticity, which are responsible for learning processes in human brains and retaining new information in neurons.

The initial devices were created at the microscale, but the team hopes to develop nanoscale versions. They are also investigating using proteins and other sugars such as those found in Aloe vera leaves, but say several properties of honey in particular shows promise.

“Honey does not spoil. It has a very low moisture concentration, so bacteria cannot survive in it. This means these computer chips will be very stable and reliable for a very long time,” said Zhao. “When we want to dispose of devices using computer chips made of honey, we can easily dissolve them in water. Because of these special properties, honey is very useful for creating renewable and biodegradable neuromorphic systems.”