Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks

Despite the outstanding performance of deep neural networks, they remain vulnerable to adversarial attacks. While digital domain adversarial attacks are well-documented, most physical-world attacks are typically visible to the human eye. Here, we present a novel invisible optical-based physical adve...

Full description

Saved in:
Bibliographic Details
Main Authors: Zvi Stein, Adir Hazan, Adrian Stern
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/7/2301
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850188287452381184
author Zvi Stein
Adir Hazan
Adrian Stern
author_facet Zvi Stein
Adir Hazan
Adrian Stern
author_sort Zvi Stein
collection DOAJ
description Despite the outstanding performance of deep neural networks, they remain vulnerable to adversarial attacks. While digital domain adversarial attacks are well-documented, most physical-world attacks are typically visible to the human eye. Here, we present a novel invisible optical-based physical adversarial attack via dazzling a CMOS camera. This attack involves using a designed light pulse sequence spatially transformed within the acquired image due to the camera’s shutter mechanism. We provide a detailed analysis of the photopic conditions required to keep the attacking light source invisible to human observers while effectively disrupting the image, thereby deceiving the DNN. The results indicate that the light source duty cycle controls the tradeoff between the attack’s success rate and the degree of concealment needed.
format Article
id doaj-art-cdbba427a1144941bd5a34826ba62b67
institution OA Journals
issn 1424-8220
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-cdbba427a1144941bd5a34826ba62b672025-08-20T02:15:54ZengMDPI AGSensors1424-82202025-04-01257230110.3390/s25072301Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural NetworksZvi Stein0Adir Hazan1Adrian Stern2School of Electrical and Computer Engineering, Ben-Gurion University of the Negev, Beer-Sheva 8410501, IsraelSchool of Electrical and Computer Engineering, Ben-Gurion University of the Negev, Beer-Sheva 8410501, IsraelSchool of Electrical and Computer Engineering, Ben-Gurion University of the Negev, Beer-Sheva 8410501, IsraelDespite the outstanding performance of deep neural networks, they remain vulnerable to adversarial attacks. While digital domain adversarial attacks are well-documented, most physical-world attacks are typically visible to the human eye. Here, we present a novel invisible optical-based physical adversarial attack via dazzling a CMOS camera. This attack involves using a designed light pulse sequence spatially transformed within the acquired image due to the camera’s shutter mechanism. We provide a detailed analysis of the photopic conditions required to keep the attacking light source invisible to human observers while effectively disrupting the image, thereby deceiving the DNN. The results indicate that the light source duty cycle controls the tradeoff between the attack’s success rate and the degree of concealment needed.https://www.mdpi.com/1424-8220/25/7/2301adversarial attackPSFrolling shutterCMOS
spellingShingle Zvi Stein
Adir Hazan
Adrian Stern
Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
Sensors
adversarial attack
PSF
rolling shutter
CMOS
title Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
title_full Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
title_fullStr Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
title_full_unstemmed Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
title_short Invisible CMOS Camera Dazzling for Conducting Adversarial Attacks on Deep Neural Networks
title_sort invisible cmos camera dazzling for conducting adversarial attacks on deep neural networks
topic adversarial attack
PSF
rolling shutter
CMOS
url https://www.mdpi.com/1424-8220/25/7/2301
work_keys_str_mv AT zvistein invisiblecmoscameradazzlingforconductingadversarialattacksondeepneuralnetworks
AT adirhazan invisiblecmoscameradazzlingforconductingadversarialattacksondeepneuralnetworks
AT adrianstern invisiblecmoscameradazzlingforconductingadversarialattacksondeepneuralnetworks