Survey of privacy preserving federated tree-based learning

The convergence of machine learning (ML) and the increasing concerns of data privacy have spurred the development of privacy-preserving techniques. Federated learning (FL) has emerged as a promising paradigm for collaborative model training across decentralized devices while keeping raw data local. However, FL is not inherently immune to privacy threats, particularly regarding the sharing of intermediate model updates. Tree-based models, such as Decision Trees, Random Forests, Extremely Randomized Trees, Isolation Forest, Gradient-Boosted Decision Trees, and eXtreme Gradient Boosting, are popular ML techniques known for their interpretability and robust performance. Integrating these tree-based models with FL presents unique challenges and opportunities for preserving privacy. This survey provides a comprehensive overview of the state-of-the-art in privacy preservation for federated tree-based learning. We review existing approaches that combine FL with privacy-enhancing technologies such as differential privacy, homomorphic encryption, and secure multiparty computation to protect sensitive information while training decision trees and their ensembles (i.e., random forests, extremely randomized trees, gradient-boosted decision trees, and extreme gradient-boosting) in a federated setting. We further analyze the challenges of privacy preservation techniques in tree-based FL, including the trade-off between privacy, utility, efficiency, and the complexity of handling distributed data structures. Finally, we identify key opportunities and outline future research directions in this evolving field.