Optical firewall for defending deep neural networks from adversarial attacks

Vladislav Kravets; Bahram Javidi; Adrian Stern

Optical firewall for defending deep neural networks from adversarial attacks

Vladislav Kravets, Bahram Javidi, Adrian Stern

Research output: Contribution to journal › Conference article › peer-review

Abstract

We overview approaches that we have recently introduced that employ optical encrypted acquisition to defend against adversarial attacks on deep learning algorithms.

Original language	English
Article number	FW5A.2
Journal	Optics InfoBase Conference Papers
State	Published - 1 Jan 2021
Event	Frontiers in Optics + Laser Science 2021, FiO+LS 2021 - Part of Frontiers in Optics, FiO 2021 - Virtual, Online, United States Duration: 1 Nov 2021 → 4 Nov 2021

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Mechanics of Materials

Cite this

@article{84352eb555d1412790df57120d2de349,

title = "Optical firewall for defending deep neural networks from adversarial attacks",

abstract = "We overview approaches that we have recently introduced that employ optical encrypted acquisition to defend against adversarial attacks on deep learning algorithms.",

author = "Vladislav Kravets and Bahram Javidi and Adrian Stern",

note = "Publisher Copyright: Frontiers in Optics / Laser Science {\textcopyright} Optica Publishing Group 2021.; Frontiers in Optics + Laser Science 2021, FiO+LS 2021 - Part of Frontiers in Optics, FiO 2021 ; Conference date: 01-11-2021 Through 04-11-2021",

year = "2021",

month = jan,

day = "1",

language = "English",

journal = "Optics InfoBase Conference Papers",

issn = "2162-2701",

}

TY - JOUR

T1 - Optical firewall for defending deep neural networks from adversarial attacks

AU - Kravets, Vladislav

AU - Javidi, Bahram

AU - Stern, Adrian

N1 - Publisher Copyright: Frontiers in Optics / Laser Science © Optica Publishing Group 2021.

PY - 2021/1/1

Y1 - 2021/1/1

N2 - We overview approaches that we have recently introduced that employ optical encrypted acquisition to defend against adversarial attacks on deep learning algorithms.

AB - We overview approaches that we have recently introduced that employ optical encrypted acquisition to defend against adversarial attacks on deep learning algorithms.

UR - http://www.scopus.com/inward/record.url?scp=85130308708&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85130308708

SN - 2162-2701

JO - Optics InfoBase Conference Papers

JF - Optics InfoBase Conference Papers

M1 - FW5A.2

T2 - Frontiers in Optics + Laser Science 2021, FiO+LS 2021 - Part of Frontiers in Optics, FiO 2021

Y2 - 1 November 2021 through 4 November 2021

ER -

Optical firewall for defending deep neural networks from adversarial attacks

Abstract

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this