Abstract
The growing use of artificial intelligence across edge devices, embedded systems, and autonomous machines has augmented concerns regarding the security of hardware, trustworthiness of data, and long-term dependability of these systems. The limitations of classical CMOS-based architectures in terms of energy efficiency, adaptability, and resistance to attack exacerbate this situation. In this landscape, neuromorphic computing may provide significant value as it pairs architectures that imitate the brain with state-of-the-art non-volatile semiconductor devices. This chapter focuses on the potential of AI-enabled neuromorphic hardware to be the basis for secure and trusted semiconductor systems based on memristors, ferroelectric field-effect transistors (FeFETs), phase-change memory (PCM), and spintronic devices. The chapter reviews device technologies but also explicitly relates intrinsic neuromorphic properties such as plasticity, stochasticity, and in-memory computation to specific hardware attack vectors including side-channel attacks, fault injection, and hardware Trojans.

This publication has 31 references indexed in Scilit: