In this paper, we design PIVOT, a new privacy-preserving method that supports outsourcing of text data for word embedding. PIVOT includes a 1-to-many mapping function for text documents that can defend against the frequency analysis attack with provable guarantee, while preserving the word context during transformation.
|Number of pages||3|
|Journal||CEUR Workshop Proceedings|
|State||Published - 1 Jan 2019|
|Event||2019 PAL: Privacy-Enhancing Artificial Intelligence and Language Technologies, PAL 2019 - Palo Alto, United States|
Duration: 25 Mar 2019 → 27 Mar 2019