ZS-PAG generates adversarial proxy samples from the forget set, projects unlearning updates into the orthogonal complement of remaining-class subspaces, and optimizes pseudo-labels with influence functions to enable zero-shot unlearning without over-unlearning.
Model inversion attacks that exploit confidence information and basic countermeasures
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Zero-Shot Machine Unlearning with Proxy Adversarial Data Generation
ZS-PAG generates adversarial proxy samples from the forget set, projects unlearning updates into the orthogonal complement of remaining-class subspaces, and optimizes pseudo-labels with influence functions to enable zero-shot unlearning without over-unlearning.