A Differentially Private Text Perturbation Method Using a Regularized Mahalanobis Metric
Authors:
- Zekun XuZXZekun Xu
- Abhinav AggarwalAAAbhinav Aggarwal
- Oluwaseyi FeyisetanOFOluwaseyi Feyisetan
- Nathanael TeissierNTNathanael Teissier
Preprint
Abstract
Balancing the privacy-utility tradeoff is a crucial requirement of many practical machine learning systems that deal with sensitive customer data. A popular approach for privacy-preserving text analysis is noise injection, in which text data is first mapped into a continuous embedding space, perturbed by sampling a spherical noise from an appropriate distribution, and then projected back to the discrete vocabulary space. While this allows the perturbation to admit the required metric differential privacy, often the utility of downstream tasks modeled on this perturbed data is low because the spherical noise does not account for the variability in the density around different words in the embedding space. In particular, words in a sparse region are likely unchanged even when the noise scale is large. %Using the global sensitivity of the mechanism can potentially add too much noise to the words in the dense regions of the embedding space, causing a high utility loss, whereas using local sensitivity can leak information through the scale of the noise added. In this paper, we propose a text perturbation mechanism based on a carefully designed regularized variant of the Mahalanobis metric to overcome this problem. For any given noise scale, this metric adds an elliptical noise to account for the covariance structure in the embedding space. This heterogeneity in the noise scale along different directions helps ensure that the words in the sparse region have sufficient likelihood of replacement without sacrificing the overall utility. We provide a text-perturbation algorithm based on this metric and formally prove its privacy guarantees. Additionally, we empirically show that our mechanism improves the privacy statistics to achieve the same level of utility as compared to the state-of-the-art Laplace mechanism.Keywords
All Related Versions
This publication has 14 references indexed in Scilit:
- Enriching Word Vectors with Subword InformationTransactions of the Association for Computational Linguistics, 2017
- A Low-Rank and Sparse Matrix Decomposition-Based Mahalanobis Distance Method for Hyperspectral Anomaly DetectionIEEE Transactions on Geoscience and Remote Sensing, 2015
- PEAS: Private, Efficient and Accurate Web Search2015
- Mahalanobis distance based on fuzzy clustering algorithm for image segmentationDigital Signal Processing, 2015
- C‐sanitized: A privacy model for document redaction and sanitizationJournal of the Association for Information Science and Technology, 2015
- Large dimensional analysis and optimization of robust shrinkage covariance matrix estimatorsJournal of Multivariate Analysis, 2014
- Knowledge-based scheme to create privacy-preserving but semantically-related queries for web search enginesInformation Sciences, 2013
- Genomic privacy and limits of individual detection in a poolNature Genetics, 2009
- h(k)‐private information retrieval from privacy‐uncooperative queryable databasesOnline Information Review, 2009
- Weaving Technology and Policy Together to Maintain ConfidentialityJournal of Law, Medicine & Ethics, 1997