Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks

TL;DR

This paper introduces Neural Attention Distillation (NAD), a novel method to effectively erase backdoor triggers from deep neural networks using minimal clean data, without harming model performance.

Contribution

NAD is a new defense framework that aligns student and teacher network attention to remove backdoors with limited clean data.

Findings

Effective against 6 state-of-the-art backdoor attacks

Removes backdoor triggers using only 5% clean data

Maintains performance on clean examples

Abstract

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.