Decision Tree-based Privacy Protection in Federated Learning: A Survey

Machine Learning (ML) has drawn considerable attention in recent years, as a new type of machine learning technology, Federated Learning (FL) conducts multi-party security collaborative training without exposing local raw data. Compared with traditional neural networks or linear models, decision tree models have a higher simplicity and interpretability. Fusing FL technology with decision tree models has great potential for performance improvement and privacy improvement. One of current issues is finding out the method of implementing the training and prediction of decision tree models in a FL environment. This survey addresses this issue and examines recent efforts to integrate federated learning and decision tree technology. We review the research results that have been implemented on the federated decision tree and consider data security a key focus of FL. This survey also discusses the issues of data privacy and security in the federated decision tree model. The main finding of this survey is providing theoretical support for the engineering of using decision trees as the underlying training model in FL.