Revisiting Data-Free Knowledge Distillation with Poisoned Teachers

Junyuan Hong, Yi Zeng, Shuyang Yu, Lingjuan Lyu, Ruoxi Jia, Jiayu Zhou
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:13199-13212, 2023.

Abstract

Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD .

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-hong23c, title = {Revisiting Data-Free Knowledge Distillation with Poisoned Teachers}, author = {Hong, Junyuan and Zeng, Yi and Yu, Shuyang and Lyu, Lingjuan and Jia, Ruoxi and Zhou, Jiayu}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {13199--13212}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/hong23c/hong23c.pdf}, url = {https://proceedings.mlr.press/v202/hong23c.html}, abstract = {Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD .} }
Endnote
%0 Conference Paper %T Revisiting Data-Free Knowledge Distillation with Poisoned Teachers %A Junyuan Hong %A Yi Zeng %A Shuyang Yu %A Lingjuan Lyu %A Ruoxi Jia %A Jiayu Zhou %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-hong23c %I PMLR %P 13199--13212 %U https://proceedings.mlr.press/v202/hong23c.html %V 202 %X Data-free knowledge distillation (KD) helps transfer knowledge from a pre-trained model (known as the teacher model) to a smaller model (known as the student model) without access to the original training data used for training the teacher model. However, the security of the synthetic or out-of-distribution (OOD) data required in data-free KD is largely unknown and under-explored. In this work, we make the first effort to uncover the security risk of data-free KD w.r.t. untrusted pre-trained models. We then propose Anti-Backdoor Data-Free KD (ABD), the first plug-in defensive method for data-free KD methods to mitigate the chance of potential backdoors being transferred. We empirically evaluate the effectiveness of our proposed ABD in diminishing transferred backdoor knowledge while maintaining compatible downstream performances as the vanilla KD. We envision this work as a milestone for alarming and mitigating the potential backdoors in data-free KD. Codes are released at https://github.com/illidanlab/ABD .
APA
Hong, J., Zeng, Y., Yu, S., Lyu, L., Jia, R. & Zhou, J.. (2023). Revisiting Data-Free Knowledge Distillation with Poisoned Teachers. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:13199-13212 Available from https://proceedings.mlr.press/v202/hong23c.html.

Related Material