ASMR: Angular Support for Malfunctioning Client Resilience in Federated Learning

Mirko Konstantin, Moritz Fuchs, Anirban Mukhopadhyay
Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, PMLR 250:754-767, 2024.

Abstract

Federated Learning (FL) allows the training of deep neural networks in a distributed andprivacy-preserving manner. However, this concept suffers from malfunctioning updatessent by the attending clients that cause global model performance degradation. Reasonsfor this malfunctioning might be technical issues, disadvantageous training data, or mali-cious attacks. Most of the current defense mechanisms are meant to require impracticalprerequisites like knowledge about the number of malfunctioning updates, which makesthem unsuitable for real-world applications. To counteract these problems, we introducea novel method called ASMR, that dynamically excludes malfunctioning clients based ontheir angular distance. Our novel method does not require any hyperparameters or knowl-edge about the number of malfunctioning clients. Our experiments showcase the detectioncapabilities of ASMR in an image classification task on a histopathological dataset, whilealso presenting findings on the significance of dynamically adapting decision boundaries.

Cite this Paper

BibTeX
@InProceedings{pmlr-v250-konstantin24a, title = {ASMR: Angular Support for Malfunctioning Client Resilience in Federated Learning}, author = {Konstantin, Mirko and Fuchs, Moritz and Mukhopadhyay, Anirban}, booktitle = {Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning}, pages = {754--767}, year = {2024}, editor = {Burgos, Ninon and Petitjean, Caroline and Vakalopoulou, Maria and Christodoulidis, Stergios and Coupe, Pierrick and Delingette, Hervé and Lartizien, Carole and Mateus, Diana}, volume = {250}, series = {Proceedings of Machine Learning Research}, month = {03--05 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v250/main/assets/konstantin24a/konstantin24a.pdf}, url = {https://proceedings.mlr.press/v250/konstantin24a.html}, abstract = {Federated Learning (FL) allows the training of deep neural networks in a distributed andprivacy-preserving manner. However, this concept suffers from malfunctioning updatessent by the attending clients that cause global model performance degradation. Reasonsfor this malfunctioning might be technical issues, disadvantageous training data, or mali-cious attacks. Most of the current defense mechanisms are meant to require impracticalprerequisites like knowledge about the number of malfunctioning updates, which makesthem unsuitable for real-world applications. To counteract these problems, we introducea novel method called ASMR, that dynamically excludes malfunctioning clients based ontheir angular distance. Our novel method does not require any hyperparameters or knowl-edge about the number of malfunctioning clients. Our experiments showcase the detectioncapabilities of ASMR in an image classification task on a histopathological dataset, whilealso presenting findings on the significance of dynamically adapting decision boundaries.} }
Endnote
%0 Conference Paper %T ASMR: Angular Support for Malfunctioning Client Resilience in Federated Learning %A Mirko Konstantin %A Moritz Fuchs %A Anirban Mukhopadhyay %B Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ninon Burgos %E Caroline Petitjean %E Maria Vakalopoulou %E Stergios Christodoulidis %E Pierrick Coupe %E Hervé Delingette %E Carole Lartizien %E Diana Mateus %F pmlr-v250-konstantin24a %I PMLR %P 754--767 %U https://proceedings.mlr.press/v250/konstantin24a.html %V 250 %X Federated Learning (FL) allows the training of deep neural networks in a distributed andprivacy-preserving manner. However, this concept suffers from malfunctioning updatessent by the attending clients that cause global model performance degradation. Reasonsfor this malfunctioning might be technical issues, disadvantageous training data, or mali-cious attacks. Most of the current defense mechanisms are meant to require impracticalprerequisites like knowledge about the number of malfunctioning updates, which makesthem unsuitable for real-world applications. To counteract these problems, we introducea novel method called ASMR, that dynamically excludes malfunctioning clients based ontheir angular distance. Our novel method does not require any hyperparameters or knowl-edge about the number of malfunctioning clients. Our experiments showcase the detectioncapabilities of ASMR in an image classification task on a histopathological dataset, whilealso presenting findings on the significance of dynamically adapting decision boundaries.
APA
Konstantin, M., Fuchs, M. & Mukhopadhyay, A.. (2024). ASMR: Angular Support for Malfunctioning Client Resilience in Federated Learning. Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 250:754-767 Available from https://proceedings.mlr.press/v250/konstantin24a.html.

Related Material