Split-and-Denoise: Protect large language model inference with local differential privacy.
Peihua MaiRan YanZhe HuangYoujia YangYan PangPublished in: CoRR (2023)
Keyphrases
- language model
- differential privacy
- differentially private
- language modeling
- privacy preserving
- document retrieval
- n gram
- information retrieval
- retrieval model
- probabilistic model
- query expansion
- test collection
- privacy preservation
- smoothing methods
- mixture model
- ad hoc information retrieval
- privacy protection
- data sharing
- translation model
- personal information
- search logs
- data privacy
- sensitive information
- feature selection
- data mining