NIPS2017 Adversarial Learning Challenge

该数据集是在一场专注于评估机器学习模型对抗恶意攻击鲁棒性的竞赛中使用的。此外，该数据集还用于评估一种基于空间变换的对抗性样本方法。在排除了非彩色图像之后，数据集规模为903张图像。其任务是进行对抗性攻击评估。

This dataset was utilized in a competition dedicated to evaluating the adversarial robustness of machine learning models against malicious attacks. Additionally, it is also employed to evaluate a spatial transformation-based adversarial sample method. After filtering out non-color images, the dataset contains a total of 903 images. The core task supported by this dataset is adversarial attack evaluation.

搜集汇总

背景与挑战

背景概述

NIPS2017 Adversarial Learning Challenge是一个专注于对抗性攻击和防御的竞赛，旨在推动对抗性示例和机器学习分类器鲁棒性的研究。竞赛包含三个子任务：非目标对抗攻击、目标对抗攻击和防御对抗攻击，参与者需开发相应的方法来解决这些任务。