Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks

About

Deep neural networks (DNNs) are known vulnerable to backdoor attacks, a training time attack that injects a trigger pattern into a small proportion of training data so as to control the model's prediction at the test time. Backdoor attacks are notably dangerous since they do not affect the model's performance on clean examples, yet can fool the model to make incorrect prediction whenever the trigger pattern appears during testing. In this paper, we propose a novel defense framework Neural Attention Distillation (NAD) to erase backdoor triggers from backdoored DNNs. NAD utilizes a teacher network to guide the finetuning of the backdoored student network on a small clean subset of data such that the intermediate-layer attention of the student network aligns with that of the teacher network. The teacher network can be obtained by an independent finetuning process on the same clean subset. We empirically show, against 6 state-of-the-art backdoor attacks, NAD can effectively erase the backdoor triggers using only 5\% clean training data without causing obvious performance degradation on clean examples. Code is available in https://github.com/bboylyg/NAD.

Yige Li, Xixiang Lyu, Nodens Koren, Lingjuan Lyu, Bo Li, Xingjun Ma• 2021

Related benchmarks

TaskDatasetResultRank
Image ClassificationMNIST (test)
Accuracy99.26
882
Backdoor DefenseCIFAR10 (test)
ASR1.6
322
Backdoor DefenseGTSRB (test)
ASR0.01
127
Backdoor DefenseTiny-ImageNet
Accuracy57.42
102
Image ClassificationGTSRB
Natural Accuracy96.53
87
Backdoor DefenseCIFAR-10
Attack Success Rate88.1
78
Image ClassificationMNIST
Clean Accuracy99.24
71
Backdoor DefenseAverage of four datasets (test)
Accuracy92.25
70
Backdoor DefenseCIFAR10 (train)
ASR1.03
63
Bias DefenseAverage of four datasets (test)
Accuracy92.17
56
Showing 10 of 36 rows

Other info

Follow for update