Differentially Private Fine-tuning of Language Models

Abstract

We give simpler, sparser, and faster algorithms for differentially private fine-tuning of large-scale pre-trained language models, which achieve the state-of-the-art privacy versus utility tradeoffs on many standard NLP tasks. We propose a meta-framework for this problem, inspired by the recent success of highly parameter-efficient methods for fine-tuning. Our experiments show that differentially private adaptations of these approaches outperform previous private algorithms in three important dimensions: utility, privacy, and the computational and memory cost of private training. On many commonly studied datasets, the utility of private models approaches that of non-private models. For example, on the MNLI dataset we achieve an accuracy of $87.8\%$ using RoBERTa-Large and $83.5\%$ using RoBERTa-Base with a privacy budget of $\epsilon =6.7$. In comparison, absent privacy constraints, RoBERTa-Large achieves an accuracy of $90.2\%$. Our findings are similar for natural language generation tasks. Privately fine-tuning with DART, GPT-2-Small, GPT-2-Medium, GPT-2-Large, and GPT-2-XL achieve BLEU scores of 38.5, 42.0, 43.1, and 43.8 respectively (privacy budget of $\epsilon =6.8,\delta =$ 1e-5) whereas the non-private baseline is $48.1$. All our experiments suggest that larger models are better suited for private fine-tuning: while they are well known to achieve superior accuracy non-privately, we find that they also better maintain their accuracy when privacy is introduced.

ICLR 2022. Code available at https://github.com/huseyinatahaninan/Differentially-Private-Fine-tuning-of-Language-Models

References in corpus (7)

• LoRA: Low-Rank Adaptation of Large Language Models
• Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data
• Large Language Models Can Be Strong Differentially Private Learners
• Training Production Language Models without Memorizing User Data
• Limits of Private Learning with Access to Public Data
• Private Query Release Assisted by Public Data
• Learning from Mixtures of Private and Public Populations

Cited by in corpus (9)

• How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy
• Privately Fine-Tuning Large Language Models with Differential Privacy
• DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Pass
• Differential Privacy in Natural Language Processing: The Story So Far
• Differential privacy and robust statistics in high dimensions
• Public Data-Assisted Mirror Descent for Private Model Training
• Modelling Direct Messaging Networks with Multiple Recipients for Cyber Deception
• Differentially Private Adapters for Parameter Efficient Acoustic Modeling
• Differential Privacy, Linguistic Fairness, and Training Data Influence: Impossibility and Possibility Theorems for Multilingual Language Models