eNVM based In-memory Computing for Intelligent and Secure Computing Systems

Kejie Huang, Chuyun Qin

Abstract: The recent breakthrough in Artificial Intelligence (AI) largely relies on the advance of hardware performance. However, energy and security have been primary concerns in the design and management of modern computing systems and data centers. The emerging embedded Non-Volatile Memories (eNVMs) have drawn great attention to various applications to improve both computing speed and security with much lower power consumption. In this paper, we review and discuss several application cases of the in-memory computing techniques, including Deep Neural Network (DNN) accelerators and neuromorphic computing circuits. Moreover, we also explore the in-memory computing techniques for the security enhancement, especially for machine learning attack. The emerging security issues raised by eNVMs and possible solutions are also discussed.