arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1702.06763cs.LGcs.AIcs.CR
收藏

DeepCloak:掩蔽深度神经网络模型以增强对抗性样本的鲁棒性

DeepCloak: Masking Deep Neural Network Models for Robustness Against Adversarial Samples

University of Virginia(弗吉尼亚大学)

Ji Gao, Beilun Wang, Zeming Lin, Weilin Xu, Yanjun Qi

摘要:

最近的研究表明,深度神经网络(DNN)很容易受到对抗性样本的攻击:恶意干扰的样本会产生错误的模型输出。这种攻击可能会严重破坏DNN系统,特别是在安全敏感的环境中。据观察,对手可以通过对当前分类任务不必要的不相关特征维度进行小扰动来轻松生成对抗样本。为了克服这个问题,我们引入了一种名为DeepCloak的防御机制。通过识别和删除DNN模型中不必要的特征,DeepCloak限制了攻击者生成对抗样本的能力,从而提高了对此类输入的鲁棒性。与其他防御方法相比,DeepCloak易于实现,计算效率高。实验结果表明,DeepCloak可以提高最先进的DNN模型对对抗性样本的性能。

英文摘要:

Recent studies have shown that deep neural networks (DNN) are vulnerable to adversarial samples: maliciously-perturbed samples crafted to yield incorrect model outputs. Such attacks can severely undermine DNN systems, particularly in security-sensitive settings. It was observed that an adversary could easily generate adversarial samples by making a small perturbation on irrelevant feature dimensions that are unnecessary for the current classification task. To overcome this problem, we introduce a defensive mechanism called DeepCloak. By identifying and removing unnecessary features in a DNN model, DeepCloak limits the capacity an attacker can use generating adversarial samples and therefore increase the robustness against such inputs. Comparing with other defensive approaches, DeepCloak is easy to implement and computationally efficient. Experimental results show that DeepCloak can increase the performance of state-of-the-art DNN models against adversarial samples.

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。

↑