Invariant Aggregator for Defending against Federated Backdoor Attacks

Xiaoyang Wang, Dimitrios Dimitriadis, Sanmi Koyejo, Shruti Tople
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:2728-2736, 2024.

Abstract

Federated learning enables training high-utility models across several clients without directly sharing their private data. As a downside, the federated setting makes the model vulnerable to various adversarial attacks in the presence of malicious clients. Despite the theoretical and empirical success in defending against attacks that aim to degrade models’ utility, defense against backdoor attacks that increase model accuracy on backdoor samples exclusively without hurting the utility on other samples remains challenging. To this end, we first analyze the failure modes of existing defenses over a flat loss landscape, which is common for well-designed neural networks such as Resnet (He et al., 2015) but is often overlooked by previous works. Then, we propose an invariant aggregator that redirects the aggregated update to invariant directions that are generally useful via selectively masking out the update elements that favor few and possibly malicious clients. Theoretical results suggest that our approach provably mitigates backdoor attacks and remains effective over flat loss landscapes. Empirical results on three datasets with different modalities and varying numbers of clients further demonstrate that our approach mitigates a broad class of backdoor attacks with a negligible cost on the model utility.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-wang24e, title = { Invariant Aggregator for Defending against Federated Backdoor Attacks }, author = {Wang, Xiaoyang and Dimitriadis, Dimitrios and Koyejo, Sanmi and Tople, Shruti}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {2728--2736}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/wang24e/wang24e.pdf}, url = {https://proceedings.mlr.press/v238/wang24e.html}, abstract = { Federated learning enables training high-utility models across several clients without directly sharing their private data. As a downside, the federated setting makes the model vulnerable to various adversarial attacks in the presence of malicious clients. Despite the theoretical and empirical success in defending against attacks that aim to degrade models’ utility, defense against backdoor attacks that increase model accuracy on backdoor samples exclusively without hurting the utility on other samples remains challenging. To this end, we first analyze the failure modes of existing defenses over a flat loss landscape, which is common for well-designed neural networks such as Resnet (He et al., 2015) but is often overlooked by previous works. Then, we propose an invariant aggregator that redirects the aggregated update to invariant directions that are generally useful via selectively masking out the update elements that favor few and possibly malicious clients. Theoretical results suggest that our approach provably mitigates backdoor attacks and remains effective over flat loss landscapes. Empirical results on three datasets with different modalities and varying numbers of clients further demonstrate that our approach mitigates a broad class of backdoor attacks with a negligible cost on the model utility. } }
Endnote
%0 Conference Paper %T Invariant Aggregator for Defending against Federated Backdoor Attacks %A Xiaoyang Wang %A Dimitrios Dimitriadis %A Sanmi Koyejo %A Shruti Tople %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-wang24e %I PMLR %P 2728--2736 %U https://proceedings.mlr.press/v238/wang24e.html %V 238 %X Federated learning enables training high-utility models across several clients without directly sharing their private data. As a downside, the federated setting makes the model vulnerable to various adversarial attacks in the presence of malicious clients. Despite the theoretical and empirical success in defending against attacks that aim to degrade models’ utility, defense against backdoor attacks that increase model accuracy on backdoor samples exclusively without hurting the utility on other samples remains challenging. To this end, we first analyze the failure modes of existing defenses over a flat loss landscape, which is common for well-designed neural networks such as Resnet (He et al., 2015) but is often overlooked by previous works. Then, we propose an invariant aggregator that redirects the aggregated update to invariant directions that are generally useful via selectively masking out the update elements that favor few and possibly malicious clients. Theoretical results suggest that our approach provably mitigates backdoor attacks and remains effective over flat loss landscapes. Empirical results on three datasets with different modalities and varying numbers of clients further demonstrate that our approach mitigates a broad class of backdoor attacks with a negligible cost on the model utility.
APA
Wang, X., Dimitriadis, D., Koyejo, S. & Tople, S.. (2024). Invariant Aggregator for Defending against Federated Backdoor Attacks . Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:2728-2736 Available from https://proceedings.mlr.press/v238/wang24e.html.

Related Material