Differential Privacy for Text Analytics via Natural Text Sanitization
This work addresses privacy concerns in text analytics for users handling sensitive data, offering an incremental improvement over existing methods.
The paper tackled the problem of low utility in differentially private text sanitization by introducing a new local DP notion that considers sensitivity and similarity, resulting in promising utility for privacy-preserving NLP with BERT without increasing inference attack success rates.
Texts convey sophisticated knowledge. However, texts also convey sensitive information. Despite the success of general-purpose language models and domain-specific mechanisms with differential privacy (DP), existing text sanitization mechanisms still provide low utility, as cursed by the high-dimensional text representation. The companion issue of utilizing sanitized texts for downstream analytics is also under-explored. This paper takes a direct approach to text sanitization. Our insight is to consider both sensitivity and similarity via our new local DP notion. The sanitized texts also contribute to our sanitization-aware pretraining and fine-tuning, enabling privacy-preserving natural language processing over the BERT language model with promising utility. Surprisingly, the high utility does not boost up the success rate of inference attacks.