DPMLM

This is the code repository for the ACL Findings paper: DP-MLM: Differentially Private Text Rewriting Using Masked Language Models

Setup

In this repository, you will find a requirements.txt file, which contains all necessary Python dependencies.

Otherwise, there are two main files, both of which arte easily importable and reusable:

• DPMLM.py: code for running the DP-MLM mechanism. privatize replaces a single token, while dpmlm_rewrite will rewrite an entire text.
• LLMDP.py: implementations of both DP-Paraphrase and DP-Prompt. Note that for DP-Prompt, you will need to download the corresponding LMs, i.e., from Hugging Face.

Usage of DP-MLM

M = DPMLM.DPMLM()

M.dpmlm_rewrite("hello world", epsilon=100)

Usage of other evaluated models

M = LLMDP.DPPrompt()

M.privatize("hello world", epsilon=100)

Important notes

In order to use LLMDP.DPParaphrase, you must download the fine-tuned model directory. This can be found at the following link: Model

Also, you will need to download the wordnet 2022 corpus: python -m wn download oewn:2022

Finally, each code implementation sets specific clipping bounds, which was done for the purposes of comparable evaluation in the paper. These can be freely changed in the parameters, and should be experimented with for (possibly) better performance.