Everything from smart phones to supercomputers needs memory, and tomorrow’s devices will demand faster, more energy-efficient memory technologies to store ever increasing data.
Non-volatile memory (NVM) is a type of storage that retains data even after the power has been turned off. While this feature eliminates the need for constant power to maintain stored data, the data left on the device is vulnerable to theft.
“What we are trying to do is to keep a malicious third party from gaining access to the data stored on this type of memory,” said Kai Ni, assistant professor of electrical engineering at the University of Notre Dame.
His research on embedding security in memory, carried out in collaboration with the lab of Vijaykrishnan Narayanan at Pennsylvania State University, was recently published in Nature Communications.
“Instead of using software to encrypt data—the method typically used with such high-security transactions as online banking—we’re embedding both encryption and decryption on the memory hardware,” said Ni.
Device-level security is particularly important with the development of the Internet of Things, which enables interconnected devices—smart thermostats, traffic management systems, medical implants—fast access to stored data without requiring battery-draining energy.
But how can you provide robust security to all these interconnected devices without sacrificing energy efficiency?
“Previous types of encryption have required energy- and time-intensive computation,” Ni said. “Our system uses XOR gates, a digital logic component, and the building blocks for that already exist in memory. To decrypt the data, we use a key—a kind of secret code dictionary—which is kept safe on the device.” With this approach, the amount of time required to perform these operations was reduced by 90 percent.
Ni, who joined the Notre Dame faculty in 2023, focuses on leveraging the uniquely versatile properties of ferroelectric materials for device development. His most recent paper, which appeared in Science Advances, proposes the use of ferroelectric transistors to enable more efficient, nimble neural nets.
— Karla Cruise, Notre Dame Engineering