From Centralized to Federated Learning: What Happens to Model Explainability?

• Daniel Ribeiro Trindade UFES / IFES
• Eduardo Zambon UFES
• Rodolfo da Silva Villaça UFES
• Diego Roberto Colombo Dias UFES
• Giovanni Comarela UFES

DOI: https://doi.org/10.5753/sbrc.2026.19824

Abstract

This paper investigates the impact of federated learning (FL) on the explainability of machine learning models, with a focus on SHAP-based explanations. A comparative methodology is proposed between centrally trained and federated models, considering scenarios with uniform and non-uniform data distributions across clients. The approach evaluates the similarity of explanations using metrics such as cosine distance, ranking similarity, and sign consistency, and is validated using the EHMS dataset for attack detection in healthcare systems. Results show that federated models produce explanations that differ from those obtained with centralized training, and data heterogeneity significantly affects the explanation consistency: in uniform scenarios, local and global explanations are consistent, whereas in non-uniform settings they diverge.

References

Bak, M. et al. (2024). Federated learning is not a cure-all for data ethics. Nature Machine Intelligence, 6.

Branson, J. et al. (2020). Evaluating the re-identification risk of a clinical study report anonymized under EMA Policy 0070 and Health Canada Regulations. Trials, 21.

Chen, P. et al. (2022). EVFL: An explainable vertical federated learning for data-oriented artificial intelligence systems. Journal of Systems Architecture, 126:102474.

Chen, X. et al. (2021). Fed-EINI: An efficient and interpretable inference framework for decision tree ensembles in federated learning.

Dong, T. et al. (2022). An interpretable federated learning-based network intrusion detection framework.

Ducange, P. et al. (2024). Consistent post-hoc explainability in federated learning through federated fuzzy clustering. In 2024 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pages 1–10.

Guidotti, R. (2022). Counterfactual explanations and how to find them: literature review and benchmarking. Data Mining and Knowledge Discovery, 38:1–55.

Hady, A. A. et al. (2020). Intrusion detection system for healthcare systems using medical and network data: A comparison study. IEEE Access, 8:106576–106584.

Haffar, R. et al. (2022). Explaining predictions and attacks in federated learning via random forests. Applied Intelligence, 53:1–17.

Hou, B. et al. (2022). Mitigating the backdoor attack by federated filters for industrial IoT applications. IEEE Transactions on Industrial Informatics, 18(5):3562–3571.

Imakura, A. et al. (2020). Interpretable collaborative data analysis on distributed data.

Kalakoti, R. et al. (2025). Federated learning of explainable AI(FedXAI) for deep learning-based intrusion detection in IoT networks. Computer Networks, 270:111479.

Li, A. et al. (2023). Towards interpretable federated learning.

Liang, Z. and Wang, H. (2022). FedTSC: a secure federated learning system for interpretable time series classification. Proc. VLDB Endow., 15(12):3686–3689.

Linardatos, P. et al. (2021). Explainable AI: A review of machine learning interpretability methods. Entropy, 23(1).

Lopez-Ramos, L. M. et al. (2024). Interplay between federated learning and explainable artificial intelligence: a scoping review.

Lundberg, S. and Lee, S.-I. (2017). A unified approach to interpreting model predictions.

Ma, X. and Gu, L. (2023). Research and application of generative-adversarial-network attacks defense method based on federated learning. Electronics, 12:975.

Malandrino, F. and Chiasserini, C. F. (2021). Toward node liability in federated learning: Computational cost and network overhead. IEEE Communications Magazine, 59(9):72–77.

Markus, A. F. et al. (2021). The role of explainability in creating trustworthy artificial intelligence for health care: A comprehensive survey of the terminology, design choices, and evaluation strategies. Journal of Biomedical Informatics, 113:103655.

McMahan, H. B. et al. (2023). Communication-efficient learning of deep networks from decentralized data.

Polato, M. et al. (2022). Boosting the federation: Cross-silo federated learning without gradient descent. In 2022 Int Joint Conf on Neural Networks (IJCNN), pages 1–10.

Sarmento, E. M. et al. (2024). MininetFed: A tool for assessing client selection, aggregation, and security in federated learning. In 2024 IEEE 10th World Forum on Internet of Things (WF-IoT), pages 1–6. IEEE.

Selvaraju, R. R. et al. (2019). Grad-CAM: Visual explanations from deep networks via gradient-based localization. International Journal of Computer Vision, 128(2):336–359.

Shapley, L. S. (1953). A Value for n-Person Games, pages 307–318. Princeton Univ. Press.

Wang, G. (2019). Interpret federated learning with Shapley values.

Yang, Q. et al., editors (2020). Federated Learning Privacy and Incentive, volume 12500 of Lecture Notes in Computer Science. Springer.

Younis, R. et al. (2023). FLAMES2Graph: An interpretable federated multivariate time series classification framework. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD ’23, page 3140–3150. ACM.

Yuan, X. et al. (2022). An efficient digital twin assisted clustered federated learning algorithm for disease prediction. In 2022 IEEE 95th Vehicular Technology Conference: (VTC2022-Spring), pages 1–6.