Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimization
About
Backdoor defense, which aims to detect or mitigate the effect of malicious triggers introduced by attackers, is becoming increasingly critical for machine learning security and integrity. Fine-tuning based on benign data is a natural defense to erase the backdoor effect in a backdoored model. However, recent studies show that, given limited benign data, vanilla fine-tuning has poor defense performance. In this work, we provide a deep study of fine-tuning the backdoored model from the neuron perspective and find that backdoorrelated neurons fail to escape the local minimum in the fine-tuning process. Inspired by observing that the backdoorrelated neurons often have larger norms, we propose FTSAM, a novel backdoor defense paradigm that aims to shrink the norms of backdoor-related neurons by incorporating sharpness-aware minimization with fine-tuning. We demonstrate the effectiveness of our method on several benchmark datasets and network architectures, where it achieves state-of-the-art defense performance. Overall, our work provides a promising avenue for improving the robustness of machine learning models against backdoor attacks.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Backdoor Defense | CIFAR10 (test) | -- | 322 | |
| Backdoor Detection | CIFAR-10 | Bd. Rate20 | 120 | |
| Backdoor Defense | CIFAR-10 (test) | -- | 40 | |
| Backdoor Detection | GTSRB | -- | 39 | |
| Backdoor Defense | CIFAR-10 Blended v1 (test) | Clean Accuracy92.28 | 34 | |
| Backdoor Defense | CIFAR-10 BadNet v1 (test) | Clean Accuracy91.53 | 20 | |
| Backdoor Defense | CIFAR-10 | BadNet C-Acc92.63 | 17 | |
| Backdoor Defense | CIFAR-10 LC DenseNet-161 (test) | Clean Accuracy87.79 | 17 | |
| Backdoor Defense | CIFAR-10 SSBA DenseNet-161 (test) | Clean Accuracy86.98 | 17 | |
| Backdoor Defense | CIFAR-10 SSBA v1 (test) | Clean Accuracy91.77 | 17 |