Differentially Private Learning Needs Better Model Initialization and Self-Distillation

• URL: http://arxiv.org/abs/2410.17566v1
• Date: Wed, 23 Oct 2024 05:19:51 GMT
• Title: Differentially Private Learning Needs Better Model Initialization and Self-Distillation
• Authors: Ivoline C. Ngong, Joseph P. Near, Niloofar Mireshghallah,
• Abstract summary: Differentially private SGD (DPSGD) enables privacy-preserving training of language models, but often reduces utility, diversity, and linguistic quality. We introduce DPRefine, a three-phase method that initializes a model using data from a small pre-trained LM with rigorous filtering. We find that small models like GPT-2 can be effective for synthesis and distillation, highlighting their potential in enabling scalable and efficient deployment of privacy-preserving language.
• Score: 1.8069913326395433
• License:
• Abstract: Differentially private SGD (DPSGD) enables privacy-preserving training of language models, but often reduces utility, diversity, and linguistic quality. We introduce DPRefine, a three-phase method that initializes a model using data synthesis from a small pre-trained LM with rigorous filtering, applies DP finetuning on private data, and performs self-distillation to refine outputs. This approach significantly outperforms vanilla DPSGD, with AlpacaEval preferring DPRefine's generations in 78.4% of cases across all datasets. Our analysis reveals that DPRefine reduces linguistic errors in generated text by 84.0%, mitigating grammar and spelling errors, commonly associated with DPSGD. It also reduces inconsistencies of non-private models, such as hallucinated details and misattributed quotes. We find that small models like GPT-2 can be effective for initialization and distillation, highlighting their potential in enabling scalable and efficient deployment of privacy-preserving language.

Related papers

• Pre-training Differentially Private Models with Limited Public Data [54.943023722114134]
  differential privacy (DP) is a prominent method to gauge the degree of security provided to the models. DP is yet not capable of protecting a substantial portion of the data used during the initial pre-training stage. We develop a novel DP continual pre-training strategy using only 10% of public data. Our strategy can achieve DP accuracy of 41.5% on ImageNet-21k, as well as non-DP accuracy of 55.7% and and 60.0% on downstream tasks Places365 and iNaturalist-2021.
  arXiv  Detail & Related papers  (2024-02-28T23:26:27Z)
• Private Fine-tuning of Large Language Models with Zeroth-order Optimization [51.19403058739522]
  Differentially private gradient descent (DP-SGD) allows models to be trained in a privacy-preserving manner. We introduce DP-ZO, a private fine-tuning framework for large language models by privatizing zeroth order optimization methods.
  arXiv  Detail & Related papers  (2024-01-09T03:53:59Z)
• Locally Differentially Private Document Generation Using Zero Shot Prompting [61.20953109732442]
  We propose a locally differentially private mechanism called DP-Prompt to counter author de-anonymization attacks. When DP-Prompt is used with a powerful language model like ChatGPT (gpt-3.5), we observe a notable reduction in the success rate of de-anonymization attacks.
  arXiv  Detail & Related papers  (2023-10-24T18:25:13Z)
• Harnessing large-language models to generate private synthetic text [18.863579044812703]
  Differentially private training algorithms like DP-SGD protect sensitive training data by ensuring that trained models do not reveal private information. This paper studies an alternative approach to generate synthetic data that is differentially private with respect to the original data, and then non-privately training a model on the synthetic data. generating private synthetic data is much harder than training a private model.
  arXiv  Detail & Related papers  (2023-06-02T16:59:36Z)
• Large Scale Transfer Learning for Differentially Private Image Classification [51.10365553035979]
  Differential Privacy (DP) provides a formal framework for training machine learning models with individual example level privacy. Private training using DP-SGD protects against leakage by injecting noise into individual example gradients. While this result is quite appealing, the computational cost of training large-scale models with DP-SGD is substantially higher than non-private training.
  arXiv  Detail & Related papers  (2022-05-06T01:22:20Z)
• Just Fine-tune Twice: Selective Differential Privacy for Large Language Models [69.66654761324702]
  We propose a simple yet effective just-fine-tune-twice privacy mechanism to achieve SDP for large Transformer-based language models. Experiments show that our models achieve strong performance while staying robust to the canary insertion attack.
  arXiv  Detail & Related papers  (2022-04-15T22:36:55Z)
• Large Language Models Can Be Strong Differentially Private Learners [70.0317718115406]
  Differentially Private (DP) learning has seen limited success for building large deep learning models of text. We show that this performance drop can be mitigated with the use of large pretrained models. We propose a memory saving technique that allows clipping in DP-SGD to run without instantiating per-example gradients.
  arXiv  Detail & Related papers  (2021-10-12T01:45:27Z)
• Accuracy, Interpretability, and Differential Privacy via Explainable Boosting [22.30100748652558]
  We show that adding differential privacy to Explainable Boosting Machines (EBMs) yields state-of-the-art accuracy while protecting privacy. Our experiments on multiple classification and regression datasets show that DP-EBM models suffer surprisingly little accuracy loss even with strong differential privacy guarantees.
  arXiv  Detail & Related papers  (2021-06-17T17:33:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.