FLTrojan: Privacy Leakage Attacks against Federated Language Models Through Selective Weight Tampering.
Md. Rafi Ur RashidVishnu Asutosh DasuKang GuNajrin SultanaShagufta MehnazPublished in: CoRR (2023)
Keyphrases
- language model
- information leakage
- watermarking scheme
- personal information
- security risks
- language modeling
- sensitive information
- n gram
- document retrieval
- detecting malicious
- test collection
- privacy preserving
- probabilistic model
- information retrieval
- speech recognition
- digital images
- query expansion
- robust watermarking scheme
- statistical language models
- language modelling
- retrieval model
- query terms
- ad hoc information retrieval
- digital libraries
- translation model
- copyright protection
- pseudo relevance feedback
- language model for information retrieval
- context sensitive
- vector space model
- watermarking algorithm
- smoothing methods
- security issues
- authentication scheme
- weighting scheme
- language models for information retrieval
- jpeg compression
- watermark embedding
- authentication protocol
- relevance model
- cross lingual
- security protocols
- statistical language modeling
- data sources