提出DECAF方法，让模型高效抹除特定数据影响且防集群泄露。
DECAF: De-Clustering for Adaptive Representational Unlearning
- 仅对要删除数据加噪、抑制置信度、分散输出，破坏特征聚类结构。
- 在CIFAR-10上遗忘类准确率仅0.10%，保留准确率达79.4%。
- 适合需随时响应数据删除请求的隐私敏感场景，效率远超全量重训练。
机器遗忘旨在移除特定训练数据对已训练模型的影响，是隐私保护、责任追溯和持续部署的关键需求。我们指出，许多现有方法易受聚类攻击，可无监督恢复类别结构，限制其在需可靠按需处理删除请求场景中的应用。为此，我们提出后处理方法DECAF（De-Clustering for Adaptive Forgetting），仅作用于待遗忘数据集，通过输入噪声、置信度抑制与基于熵的输出多样化，破坏与遗忘数据相关的残差特征空间结构。在使用ResNet-18的CIFAR-10上，DECAF实现0.10%遗忘类准确率、79.4%保留准确率及0.88的AUS，优于所有基线方法。聚类分析显示，其性能接近需完整训练集的方法，但效率显著提升。代码已开源。
原文摘要 · Abstract (English)
Machine unlearning, which aims to remove the influence of specific training data from a trained model, is a key requirement for privacy, accountability, and adaptive deployment. We argue that many unlearning methods are vulnerable to a simple clustering attack, which can recover class structure in an unsupervised manner, limiting their suitability for continual deployment where removal requests must be handled reliably on demand. To address this, we propose DECAF (DE-Clustering for Adaptive Forgetting), a post-hoc method that operates only on the forget set and is designed to break the cluster. DECAF combines input noise, confidence suppression, and entropy-based output diversification to disrupt the residual feature-space structure associated with forgotten data. On CIFAR-10 with ResNet-18, DECAF attains 0.10% forget-class accuracy, 79.4% retain accuracy, and an AUS of 0.88, surpassing all other baselines. In cluster-based analysis, it attains performance comparable to that of unlearning methods that use the full training set, while being significantly more efficient. Code: https://github.com/ale256/representation_unlearning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。