Hardware Accelerator for Adversarial Attacks on Deep Learning Neural Networks

By H. Guo et al

Read the original document by opening this link in a new tab.

Table of Contents

Abstract, I. Introduction, II. Background, III. A3 Architecture, IV. Pipeline Analysis

Summary

Recent studies identify that Deep learning Neural Networks (DNNs) are vulnerable to subtle perturbations, which can fool the DNN models and lead to wrong outputs. In this paper, a hardware accelerator for adversarial attacks based on memristor crossbar arrays is proposed to improve the robustness and security of future deep learning systems. The paper discusses the architecture of the accelerator, trade-offs between storage and compute engines, and techniques to improve crossbar utilization. The study demonstrates the effectiveness of the proposed accelerator design.