arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1511.07528cs.CRcs.LGcs.NEstat.ML
收藏

深度学习在对抗环境中的局限性

The Limitations of Deep Learning in Adversarial Settings

Penn State University(宾夕法尼亚州立大学);University of Wisconsin-Madison(威斯康星大学麦迪逊分校);Carnegie Mellon University(卡内基梅隆大学);United States Army Research Laboratory(美国陆军研究实验室)

Nicolas Papernot, Patrick McDaniel, Somesh Jha, Matt Fredrikson, Z. Berkay Celik, Ananthram Swami

摘要:

深度学习利用大型数据集和计算效率高的训练算法,在各种机器学习任务中优于其他方法。然而,深度神经网络训练阶段的缺陷使它们容易受到对抗性样本的攻击:对手精心制作的输入,目的是导致深度神经网络错误分类。在这项工作中,我们形式化了针对深度神经网络(DNN)的对手空间,并引入了一类新的算法,以基于对DNN输入和输出之间映射的精确理解来制作对抗样本。在计算机视觉的应用中,我们证明了我们的算法可以可靠地产生由人类主体正确分类的样本,但在特定目标中被DNN错误分类,对抗成功率为97%,而每个样本平均只修改4.02%的输入特征。然后,我们评估不同的样本类对抗扰动的脆弱性,通过定义一个硬度的措施。最后,我们描述了通过定义良性输入和目标分类之间的距离的预测度量来概述对抗性样本的防御的初步工作。

英文摘要:

Deep learning takes advantage of large datasets and computationally efficient training algorithms to outperform other approaches at various machine learning tasks. However, imperfections in the training phase of deep neural networks make them vulnerable to adversarial samples: inputs crafted by adversaries with the intent of causing deep neural networks to misclassify. In this work, we formalize the space of adversaries against deep neural networks (DNNs) and introduce a novel class of algorithms to craft adversarial samples based on a precise understanding of the mapping between inputs and outputs of DNNs. In an application to computer vision, we show that our algorithms can reliably produce samples correctly classified by human subjects but misclassified in specific targets by a DNN with a 97% adversarial success rate while only modifying on average 4.02% of the input features per sample. We then evaluate the vulnerability of different sample classes to adversarial perturbations by defining a hardness measure. Finally, we describe preliminary work outlining defenses against adversarial samples by defining a predictive measure of distance between a benign input and a target classification.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑