arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2112.03350cs.CRcs.LG
收藏

中毒深度神经网络后门触发器的测试时检测

Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks

Pennsylvania State University(宾夕法尼亚州立大学)

Xi Li, Zhen Xiang, David J. Miller, George Kesidis

摘要:

后门(特洛伊木马)攻击是针对深度神经网络(DNN)的新兴威胁。每当来自任何源类的测试样本嵌入后门模式时,被攻击的DNN将预测攻击者想要的目标类;同时正确分类干净(无攻击)的测试样本。现有的后门防御在检测DNN是否受到攻击以及在“后训练”机制中对后门模式进行逆向工程方面取得了成功:防御者可以访问要检查的DNN和独立收集的小而干净的数据集,但无法访问DNN的(可能中毒的)训练集。然而,这些防御既不能在触发后门映射的行为中抓住罪犯,也不能在测试时减轻后门攻击。在本文中,我们提出了一个“飞行中”防御后门攻击的图像分类,1)检测使用后门触发器在测试时,和2)推断类的起源(源类)检测到的触发器的例子。我们的防御的有效性进行了实验证明,对不同的强大的后门攻击。

英文摘要:

Backdoor (Trojan) attacks are emerging threats against deep neural networks (DNN). A DNN being attacked will predict to an attacker-desired target class whenever a test sample from any source class is embedded with a backdoor pattern; while correctly classifying clean (attack-free) test samples. Existing backdoor defenses have shown success in detecting whether a DNN is attacked and in reverse-engineering the backdoor pattern in a "post-training" regime: the defender has access to the DNN to be inspected and a small, clean dataset collected independently, but has no access to the (possibly poisoned) training set of the DNN. However, these defenses neither catch culprits in the act of triggering the backdoor mapping, nor mitigate the backdoor attack at test-time. In this paper, we propose an "in-flight" defense against backdoor attacks on image classification that 1) detects use of a backdoor trigger at test-time; and 2) infers the class of origin (source class) for a detected trigger example. The effectiveness of our defense is demonstrated experimentally against different strong backdoor attacks.

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。

↑