Indigenous back-end-of-line compatible SiO2-based One-Time Programmable Memory for Secured Spiking Neural Network Inference Accelerator

The low-power, biology-inspired spiking neural networks (SNN) based architectures have facilitated artificial intelligence applications for edge applications. However, the data and model security of edge devices are significant concerns. This paper proposes a back-end-of-line compatible SiO 2 -based one-time programmable memory (OTPM) as a secured memory to store the synapses for in-memory computation (IMC) to accelerate the SNN inference. The measured characteristics of an indigenous OTPM 14 × 4 array are presented, and the SNN system performance is evaluated with these measured OTPM characteristics.