TY - GEN
T1 - mFI-PSO
T2 - 2022 International Joint Conference on Neural Networks, IJCNN 2022
AU - Shu, Hai
AU - Shi, Ronghua
AU - Jia, Qiran
AU - Zhu, Hongtu
AU - Chen, Ziqi
N1 - Funding Information:
Dr. Ziqi Chen’s work is partially supported by National Natural Science Foundation of China (Grant No. 11871477) and Natural Science Foundation of Shanghai (Grant No. 21ZR1418800). Dr. Hai Shu’s work is partially supported by a startup fund from New York University.
Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Deep neural networks (DNNs) have achieved great success in image classification, but can be very vulnerable to adversarial attacks with small perturbations to images. To improve adversarial image generation for DNNs, we develop a novel method, called mFI-PSO, which utilizes a Manifold-based First-order Influence measure for vulnerable image and pixel selection and the Particle Swarm Optimization for various objective functions. Our mFI-PSO can thus effectively design adversarial images with flexible, customized options on the number of perturbed pixels, the misclassification probability, and the targeted incorrect class. Experiments demonstrate the flexibility and effectiveness of our mFI-PSO in adversarial attacks and its appealing advantages over some popular methods.
AB - Deep neural networks (DNNs) have achieved great success in image classification, but can be very vulnerable to adversarial attacks with small perturbations to images. To improve adversarial image generation for DNNs, we develop a novel method, called mFI-PSO, which utilizes a Manifold-based First-order Influence measure for vulnerable image and pixel selection and the Particle Swarm Optimization for various objective functions. Our mFI-PSO can thus effectively design adversarial images with flexible, customized options on the number of perturbed pixels, the misclassification probability, and the targeted incorrect class. Experiments demonstrate the flexibility and effectiveness of our mFI-PSO in adversarial attacks and its appealing advantages over some popular methods.
KW - adversarial attack
KW - influence measure
KW - particle swarm optimization
KW - perturbation manifold
UR - http://www.scopus.com/inward/record.url?scp=85140778487&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85140778487&partnerID=8YFLogxK
U2 - 10.1109/IJCNN55064.2022.9892433
DO - 10.1109/IJCNN55064.2022.9892433
M3 - Conference contribution
AN - SCOPUS:85140778487
T3 - Proceedings of the International Joint Conference on Neural Networks
BT - 2022 International Joint Conference on Neural Networks, IJCNN 2022 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 18 July 2022 through 23 July 2022
ER -