SecureCut: Federated Gradient Boosting Decision Trees with Efficient Machine Unlearning

TL;DR

SecureCut introduces a novel federated GBDT framework that efficiently supports data removal, including instance and feature unlearning, in vertical federated learning, ensuring privacy, utility, and compliance with data legislation.

Contribution

It is the first framework to enable machine unlearning in VFL, supporting both instance and feature removal without retraining, while maintaining model performance.

Findings

Outperforms state-of-the-art unlearning methods in utility and forgetfulness.

Effectively supports feature and instance unlearning in VFL.

Reduces retraining costs while preserving model accuracy.

Abstract

In response to legislation mandating companies to honor the \textit{right to be forgotten} by erasing user data, it has become imperative to enable data removal in Vertical Federated Learning (VFL) where multiple parties provide private features for model training. In VFL, data removal, i.e., \textit{machine unlearning}, often requires removing specific features across all samples under privacy guarentee in federated learning. To address this challenge, we propose \methname, a novel Gradient Boosting Decision Tree (GBDT) framework that effectively enables both \textit{instance unlearning} and \textit{feature unlearning} without the need for retraining from scratch. Leveraging a robust GBDT structure, we enable effective data deletion while reducing degradation of model performance. Extensive experimental results on popular datasets demonstrate that our method achieves superior model utility and forgetfulness compared to \textit{state-of-the-art} methods. To our best knowledge, this is the first work that investigates machine unlearning in VFL scenarios.