Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Jul 24, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

Backdoor Attack on Deep Neural Networks Triggered by Fault Injection Attack on Image Sensor Interface.

Tatsuya Oyama¹, Shunsuke Okura², Kota Yoshida²

¹Graduate School of Science and Engineering, Ritsumeikan University, 1-1-1 Noji-higashi, Kusatsu 525-8577, Shiga, Japan.

Sensors (Basel, Switzerland)

|July 11, 2023

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Positional identification of a candidate gene for MALE STERILITY 2 (MS2) by linkage mapping and transcriptomic data in Cryptomeria japonica D. Don.

BMC genomics·2026

Same author

Influence of Ultrafine Bubbles (UFBs) on Cadmium Toxicity in Polypedilum nubifer (Chironomidae).

Environmental analysis, health and toxicology·2026

Same author

Development of CAR T cells Targeting a Surface RNA Binding Protein for the Treatment of Acute Leukemias.

Cancer discovery·2026

Same author

Compression-Efficient Feature Extraction Method for a CMOS Image Sensor.

Sensors (Basel, Switzerland)·2026

Same author

A novel enzyme-based strategy for improving compost quality using extracts from spent mushroom substrate.

Journal of environmental management·2026

Same author

RAF-independent MEK mutations drive refractory histiocytic neoplasms but respond to ERK inhibition.

Cancer cell·2025

This study introduces a novel method for backdoor attacks on deep neural networks (DNNs) using fault injection on mobile industry processor interfaces (MIPI). This technique enhances adversarial mark generation for more stable and successful DNN backdoor attacks.

Area of Science:

Computer Science
Cybersecurity
Artificial Intelligence

Background:

Backdoor attacks induce deep neural network (DNN) misclassification using adversarial marks.
Conventional methods for creating adversarial marks are unstable due to variations in shooting environments.

Purpose of the Study:

To propose a stable method for generating adversarial marks for backdoor attacks.
To investigate the effectiveness of fault injection on the mobile industry processor interface (MIPI) for backdoor attacks.

Main Methods:

Developed an image tampering model to generate adversarial marks via fault injection on the MIPI.
Trained a backdoor model using poison data images generated by the simulation model.
Conducted backdoor attack experiments on a trained DNN model.

Keywords:

backdoor attack fault injection attack image sensor interface mobile industry processor interface (MIPI)

More Related Videos

DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning

DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning

Published on: May 10, 2024

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Related Experiment Videos

Last Updated: Jul 24, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning

DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning

Published on: May 10, 2024

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Main Results:

The backdoor model achieved 91% clean data accuracy in normal operation.
The fault injection backdoor attack achieved an 83% success rate.
The proposed method demonstrated a stable and effective approach to backdoor attacks.

Conclusions:

Fault injection on MIPI provides a stable and effective method for generating adversarial marks.
This technique significantly enhances the success rate of backdoor attacks on DNNs.
The findings highlight vulnerabilities in DNNs related to image sensor interfaces.