DP-BART for Privatized Text Rewriting under Local Differential Privacy

TL;DR

The paper introduces DP-BART, a novel system for privatized text rewriting under local differential privacy, which reduces noise and improves performance over existing methods through innovative techniques.

Contribution

DP-BART employs a new clipping method, iterative pruning, and internal representation training to enhance privacy guarantees and performance in privatized text rewriting.

Findings

Outperforms existing LDP text rewriting systems

Reduces noise required for differential privacy guarantees

Effective across multiple textual datasets and tasks

Abstract

Privatized text rewriting with local differential privacy (LDP) is a recent approach that enables sharing of sensitive textual documents while formally guaranteeing privacy protection to individuals. However, existing systems face several issues, such as formal mathematical flaws, unrealistic privacy guarantees, privatization of only individual words, as well as a lack of transparency and reproducibility. In this paper, we propose a new system 'DP-BART' that largely outperforms existing LDP systems. Our approach uses a novel clipping method, iterative pruning, and further training of internal representations which drastically reduces the amount of noise required for DP guarantees. We run experiments on five textual datasets of varying sizes, rewriting them at different privacy guarantees and evaluating the rewritten texts on downstream text classification tasks. Finally, we thoroughly discuss the privatized text rewriting approach and its limitations, including the problem of the strict text adjacency constraint in the LDP paradigm that leads to the high noise requirement.