Saved in:
Bibliographic Details
Main Authors: Liu, Zhenyu, Gagnon, Garrett, Venkataramani, Swagath, Liu, Liu
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2402.04325
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916117475229696
author Liu, Zhenyu
Gagnon, Garrett
Venkataramani, Swagath
Liu, Liu
author_facet Liu, Zhenyu
Gagnon, Garrett
Venkataramani, Swagath
Liu, Liu
contents Deep Neural Networks (DNNs) have revolutionized a wide range of industries, from healthcare and finance to automotive, by offering unparalleled capabilities in data analysis and decision-making. Despite their transforming impact, DNNs face two critical challenges: the vulnerability to adversarial attacks and the increasing computational costs associated with more complex and larger models. In this paper, we introduce an effective method designed to simultaneously enhance adversarial robustness and execution efficiency. Unlike prior studies that enhance robustness via uniformly injecting noise, we introduce a non-uniform noise injection algorithm, strategically applied at each DNN layer to disrupt adversarial perturbations introduced in attacks. By employing approximation techniques, our approach identifies and protects essential neurons while strategically introducing noise into non-essential neurons. Our experimental results demonstrate that our method successfully enhances both robustness and efficiency across several attack scenarios, model architectures, and datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2402_04325
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enhance DNN Adversarial Robustness and Efficiency via Injecting Noise to Non-Essential Neurons
Liu, Zhenyu
Gagnon, Garrett
Venkataramani, Swagath
Liu, Liu
Machine Learning
Artificial Intelligence
Cryptography and Security
Deep Neural Networks (DNNs) have revolutionized a wide range of industries, from healthcare and finance to automotive, by offering unparalleled capabilities in data analysis and decision-making. Despite their transforming impact, DNNs face two critical challenges: the vulnerability to adversarial attacks and the increasing computational costs associated with more complex and larger models. In this paper, we introduce an effective method designed to simultaneously enhance adversarial robustness and execution efficiency. Unlike prior studies that enhance robustness via uniformly injecting noise, we introduce a non-uniform noise injection algorithm, strategically applied at each DNN layer to disrupt adversarial perturbations introduced in attacks. By employing approximation techniques, our approach identifies and protects essential neurons while strategically introducing noise into non-essential neurons. Our experimental results demonstrate that our method successfully enhances both robustness and efficiency across several attack scenarios, model architectures, and datasets.
title Enhance DNN Adversarial Robustness and Efficiency via Injecting Noise to Non-Essential Neurons
topic Machine Learning
Artificial Intelligence
Cryptography and Security
url https://arxiv.org/abs/2402.04325