Private Text Classification

TL;DR

This paper investigates privacy-preserving methods for text classification, employing techniques like homomorphic encryption and secure computation to enable analysis on confidential text data without compromising privacy.

Contribution

It introduces a framework combining various privacy-preserving techniques for text analytics and develops preliminary methods for binary classifiers on private text corpora.

Findings

Demonstrated the feasibility of privacy-preserving text classification.

Developed initial approaches for binary classifiers using Rademacher operators.

Showed how to adapt text processing to privacy-preserving techniques.

Abstract

Confidential text corpora exist in many forms, but do not allow arbitrary sharing. We explore how to use such private corpora using privacy preserving text analytics. We construct typical text processing applications using appropriate privacy preservation techniques (including homomorphic encryption, Rademacher operators and secure computation). We set out the preliminary materials from Rademacher operators for binary classifiers, and then construct basic text processing approaches to match those binary classifiers.