Análise do Aprendizado Federado em Redes Móveis

• Kaylani Bochie UFRJ
• Matteo Sammarco AXA
• Marcin Detyniecki AXA
• Miguel Elias M. Campista UFRJ

DOI: https://doi.org/10.5753/sbrc.2021.16712

Abstract

The increase of personal data collection for service customization threatens users' privacy. In the context of federated learning, privacy can be preserved by distributing the learning process, where only neural networks' weights are shared between clients and servers. This work evaluates the impact of different computer networks' parameters on the performance of federated learning models in a mobile oriented scenario. For this, we consider the performance of convolutional neural networks for image classification. The experiments carried out use the Flower framework and the CIFAR-10 dataset for image classification. Common parameters in mobile networks such as latency, connectivity, data volume, and client availability are evaluated. The results indicate that, in addition to the increase in total training time, an increase in the number of disconnected users in a training round can even reduce the performance of the federated model. These results indicate the need for client-server orchestration to perform dynamic adaptation of network conditions.

References

Asad et al. (2021). Evaluating the communication efficiency in federated learning algorithms. Em 2021 IEEE 24th International Conference on Computer Supported Cooperative Work in Design (CSCWD), p. 552–557.

Beutel et al. (2020). Flower: A friendly federated learning research framework. ArXiv, abs/2007.14390.

Bochie et al. (2020). Aprendizado profundo em redes desafiadoras: Conceitos e aplicações. Em Minicursos do XXXVIII Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos (SBRC).

Brownlee, J. (2018). Better Deep Learning. Jason Brownlee, v1.8 edição. Cunha Neto, H., Menezes, D. e Fernandes, N. (2020). Privacidade do Usuário em Aprendizado Colaborativo: Federated Learning, da Teoria à Prática, capítulo 3. Sociedade Brasileira de Computação (SBC).

Dwork, C. e Roth, A. (2014). The algorithmic foundations of differential privacy. Foundations and Trends in Theoretical Computer Science, 9(3–4):211–407.

Hernandez Marcano et al. (2019). On fully homomorphic encryption for privacypreserving deep learning. Em 2019 IEEE Globecom Workshops (GC Wkshps), p. 1–6.

Kairouz et al. (2019). Advances and open problems in federated learning. Em Foundations and Trends in Machine Learning.

Krizhevsky, A. (2012). Learning multiple layers of features from tiny images. University of Toronto.

LeCun, Y., Bengio, Y. e Hinton, G. (2015). Deep learning. nature, 521(7553):436–444.

Li, M., Soltanolkotabi, M. e Oymak, S. (2020). Gradient descent with early stopping is provably robust to label noise for overparameterized neural networks. Em Chiappa, S. e Calandra, R., editores, Proceedings of the Twenty Third International Conference on Articial Intelligence and Statistics, volume 108 of Proceedings of Machine Learning Research, p. 4313–4324. PMLR.

Li et al. (2019). Privacy-preserving federated brain tumour segmentation. Em Suk, H.I., Liu, M., Yan, P. e Lian, C., editores, Machine Learning in Medical Imaging, p. 133–141, Cham. Springer International Publishing.

Lim et al. (2020). Federated learning in mobile edge networks: A comprehensive survey. IEEE Communications Surveys & Tutorials, 22(3):2031–2063.

McMahan et al. (2017). Communication-efficient learning of deep networks from decentralized data. Em Singh, A. e Zhu, J., editores, Proceedings of the 20th International Conference on Articial Intelligence and Statistics, volume 54 of Proceedings of Machine Learning Research, p. 1273–1282, Fort Lauderdale, FL, USA. PMLR.

Narayanan, A. e Shmatikov, V. (2008). Robust de-anonymization of large sparse datasets. Em 2008 IEEE Symposium on Security and Privacy (sp 2008), p. 111–125.

Nguyen, P. Q. (2011). Breaking fully-homomorphic-encryption challenges. Em Lin, D., Tsudik, G. e Wang, X., editores, Cryptology and Network Security, p. 13–14, Berlin, Heidelberg. Springer Berlin Heidelberg.

Nguyen et al. (2019). D¨Iot: A federated self-learning anomaly detection system for iot. Em 2019 IEEE 39th International Conference on Distributed Computing Systems (ICDCS), p. 756–767.

Roth et al. (2020). Federated learning for breast density classification: A real-world implementation. Em Albarqouni, S., Bakas, S., Kamnitsas, K., Cardoso, M. J., Landman, B., Li, W., Milletari, F., Rieke, N., Roth, H., Xu, D. e Xu, Z., editores, Domain Adaptation and Representation Transfer, and Distributed and Collaborative Learning, p. 181–191, Cham. Springer International Publishing.

Shokri, R. e Shmatikov, V. (2015). Privacy-preserving deep learning. Em Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security, CCS ’15, p. 1310–1321, New York, NY, USA. Association for Computing Machinery.

Simonyan, K. e Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. Em Bengio, Y. e LeCun, Y., editores, 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings.

Tran et al. (2019). Federated learning over wireless networks: Optimization model design and analysis. Em IEEE INFOCOM 2019 IEEE Conference on Computer Communications, p. 1387–1395.

Yang et al. (2020). Scheduling policies for federated learning in wireless networks. IEEE Transactions on Communications, 68(1):317–333.