5 Papers
Nan Zhou is an academic researcher from University of Science and Technology of China. The author has contributed to research in topics: Computer science & Adversarial system. The author has an hindex of 2, co-authored 4 publications.
Chat about Author
Papers
Genetic Algorithm with Multiple Fitness Functions for Generating Adversarial Examples
Chenwang Wu,Wenjian Luo,Nan Zhou,Peilan Xu,Tao Zhu +4 more
- 28 Jun 2021
TL;DR: Zhang et al. as discussed by the authors proposed a GA with multiple fitness functions (MF-GA), which divides the evolution process into three stages, i.e., exploration stage, exploitation stage, and stable stage.
27
Generating Multi-label Adversarial Examples by Linear Programming
Nan Zhou,Wenjian Luo,Xin Lin,Peilan Xu,Zhenya Zhang +4 more
- 19 Jul 2020
TL;DR: This study has proposed a novel algorithm that generates effective multi-label adversarial examples by solving a linear programming problem (MLA-LP), which minimize the l∞ norm of distortion while constraining the changes in the label loss of the example after being perturbed.
27
Hiding All Labels for Multi-label Images: An Empirical Study of Adversarial Examples
Nan Zhou,Wenjian Luo,Jiajia Zhang,Linghao Kong,Hongwei Zhang +4 more
- 18 Jul 2021
TL;DR: In this article, an empirical study of generating a multi-label adversarial example to hide all labels in a multilabel example is presented, where the objective is to make deep learning models know nothing about the environments.
14
Detecting adversarial examples by positive and negative representations
TL;DR: Zhang et al. as mentioned in this paper proposed a positive-negative classifier (PNClassifier) which is trained by both the original examples (called positive representations) and their negative representations with the same structural and semantic features.
•Posted Content
Random Directional Attack for Fooling Deep Neural Networks.
TL;DR: This paper proposes a random directed attack (RDA) for generating adversarial examples, which can attack without any internal knowledge of the model, and its performance under black-box attack is similar to that of the white- box attack in most cases, which is difficult to achieve using existing gradient-based attack methods.