DP-BART for Privatized Text Rewriting under Local Differential Privacy
This work addresses privacy protection in text sharing for sensitive domains, offering a more efficient solution compared to prior flawed methods, though it remains incremental within the LDP framework.
The paper tackles the problem of privatized text rewriting under local differential privacy (LDP) by proposing DP-BART, which outperforms existing systems by reducing noise requirements through novel clipping, iterative pruning, and training methods, achieving improved performance on downstream text classification tasks across five datasets.
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.