DP-FROST: Differentially Private Fine-tuning of Pre-trained Models with Freezing Model Parameters

Daeyoung Hong, Woohwan Jung, Kyuseok Shim


Abstract
Training models with differential privacy has received a lot of attentions since differential privacy provides theoretical guarantee of privacy preservation. For a task in a specific domain, since a large-scale pre-trained model in the same domain contains general knowledge of the task, using such a model requires less effort in designing and training the model. However, differentially privately fine-tuning such models having a large number of trainable parameters results in large degradation of utility. Thus, we propose methods that effectively fine-tune the large-scale pre-trained models with freezing unimportant parameters for downstream tasks while satisfying differential privacy. To select the parameters to be fine-tuned, we propose several efficient methods based on the gradients of model parameters. We show the effectiveness of the proposed method by performing experiments with real datasets.
Anthology ID:
2025.coling-main.465
Volume:
Proceedings of the 31st International Conference on Computational Linguistics
Month:
January
Year:
2025
Address:
Abu Dhabi, UAE
Editors:
Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, Steven Schockaert
Venue:
COLING
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
6966–6984
Language:
URL:
https://aclanthology.org/2025.coling-main.465/
DOI:
Bibkey:
Cite (ACL):
Daeyoung Hong, Woohwan Jung, and Kyuseok Shim. 2025. DP-FROST: Differentially Private Fine-tuning of Pre-trained Models with Freezing Model Parameters. In Proceedings of the 31st International Conference on Computational Linguistics, pages 6966–6984, Abu Dhabi, UAE. Association for Computational Linguistics.
Cite (Informal):
DP-FROST: Differentially Private Fine-tuning of Pre-trained Models with Freezing Model Parameters (Hong et al., COLING 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.coling-main.465.pdf