SecureCut: Federated Gradient Boosting Decision Trees with Efficient Machine Unlearning

In response to legislation mandating companies to honor the $\mathit{\text{right}}\mathit{\text{to}}\mathit{\text{be}}\mathit{\text{forgotten}}$ 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., $\mathit{\text{machine}}\mathit{\text{unlearning}}$, 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 $\mathit{\text{instance}}\mathit{\text{unlearning}}$ and $\mathit{\text{feature}}\mathit{\text{unlearning}}$ 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 $\mathit{\text{state-of-the-art}}$ methods. To our best knowledge, this is the first work that investigates machine unlearning in VFL scenarios.