@inproceedings{26841de61b3245288fed06fbf258a651,
title = "Characterizing Adversarial Samples of Convolutional Neural Networks",
abstract = "Adversarial samples aim to make deep convolutional neural networks predict incorrectly under small perturbations. This paper investigates non-targeted adversarial samples of convolutional neural networks and makes a primitive attempt to characterize adversarial samples. Two observations are made: first, adversarial perturbations are mainly in the high-frequency domain; second, adversarial categories usually have strong semantic relevance to the original categories. Our two observations provide a solid basis to understand the behavior of convolutional neural networks and thus to improve their robustness against adversarial samples.",
author = "Cheng Jiang and Qiyang Zhao and Yuzhong Liu",
note = "Publisher Copyright: {\textcopyright} 2018 IEEE.; 11th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, CISP-BMEI 2018 ; Conference date: 13-10-2018 Through 15-10-2018",
year = "2018",
month = jul,
day = "2",
doi = "10.1109/CISP-BMEI.2018.8633182",
language = "英语",
series = "Proceedings - 2018 11th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, CISP-BMEI 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
editor = "Wei Li and Qingli Li and Lipo Wang",
booktitle = "Proceedings - 2018 11th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, CISP-BMEI 2018",
address = "美国",
}