Identifies output label space as a privacy side-channel in DP continual learning, formalizes DP for CL, and demonstrates two mitigation methods yielding higher accuracy than prior work.
Identifying and eliminating csam in generative ml training data and models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Privacy Leakage via Output Label Space and Differentially Private Continual Learning
Identifies output label space as a privacy side-channel in DP continual learning, formalizes DP for CL, and demonstrates two mitigation methods yielding higher accuracy than prior work.