Detecting scene-plausible perceptible backdoors in trained DNNs without access to the training set

DOI10.1162/NECO_A_01376MaRDI QIDQ5004356zbMATH OpenOpenAlexFDO

Authors Zhen Xiang, Hang Wang, George Kesidis, David J. Miller

Publication date 30 July 2021

Published in Neural Computation (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1162/neco_a_01376

Mathematics Subject Classification ID

Artificial intelligence (68Txx)

Recommendations

• Verifying Neural Networks Against Backdoor Attacks
• Training set camouflage
• Black-box adversarial attacks by manipulating image attributes
• Machine learning in adversarial environments
• A robust generative classifier against transfer attacks based on variational auto-encoders

Cites work

• Adversarial machine learning
• Pattern classification.
• The Influence Curve and Its Role in Robust Estimation

Cited in

(1)

• Verifying Neural Networks Against Backdoor Attacks

Describes a project that uses

Uses Software

• AlexNet
• DeepFool
• ImageNet
• Grad-CAM

This page was built for publication: Detecting scene-plausible perceptible backdoors in trained DNNs without access to the training set

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5004356)