Towards Privacy-Guaranteed Label Unlearning in Vertical Federated Learning: Few-Shot Forgetting without Disclosure
面向垂直联邦学习中隐私保障的标签撤销学习:无需披露的少样本遗忘
机构 * WeBank AI Lab(WeBank人工智能实验室) ; Universiti Malaya(马来亚大学)
专题命中 医学数据与评测 :MRI(abstract);分类 cs.CV、cs.LG
AI总结 本文提出一种面向垂直联邦学习中隐私保障的标签撤销学习方法,通过流形混合机制生成合成嵌入,实现高效标签撤销与性能恢复。
Comments Accepted at ICLR2026. This paper introduces the first method for label unlearning in vertical federated learning (VFL), focused on preventing label leakage by the active party