Proposes a self-aware unlearning method inspired by hypnopaedia that uses model inversion and hypothesis testing to detect and detach backdoor triggers from machine learning models.
Model agnostic defence against backdoor attacks in machine learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Hypnopaedia-Aware Machine Unlearning via Psychometrics of Artificial Mental Imagery
Proposes a self-aware unlearning method inspired by hypnopaedia that uses model inversion and hypothesis testing to detect and detach backdoor triggers from machine learning models.