Guided diffusion improves privacy and quality of synthetic text generation
Guided Diffusion for Trustworthy Synthetic Text Generation
Cryptography and SecurityMachine Learning
Summary
Training large language models on real data risks exposing sensitive information. To address this, the authors propose a method called Guided Diffusion that generates fake but realistic text while protecting privacy. Their technique controls the text generation process to keep the meaning accurate and applies formal privacy guarantees. Experiments show it creates higher-quality synthetic text than previous methods and prevents data leaks more effectively, even working faster in some cases. They also add a new privacy defense against data poisoning and memorization.
What this means in practice
- •For data privacy engineers: Create synthetic datasets that maintain privacy guarantees while preserving data usefulness for model training and analysis.
- •For content generation teams: Produce fluent, semantically accurate synthetic text that mimics private corpora without risking exposure of sensitive information.
Authors
Qi Pang, Hanshen Xiao, Andre Manoel, Lipika Ramaswamy, Wenting Zheng, Virginia Smith, G. Edward Suh
Abstract
The rapid evolution of Large Language Models is fueled by massive datasets, yet this reliance poses severe privacy risks, as models can inadvertently memorize and even directly output sensitive training data. Synthetic text generation offers a promising solution to this challenge; however, existing methods struggle to reconcile rigorous privacy guarantees with high data utility. In this paper, we propose a new paradigm for private synthetic text generation: Guided Diffusion. Leveraging the unique steerability of Diffusion Language Models, we introduce novel sampling and re-masking procedures that actively guide the iterative denoising steps. These mechanisms steer the generative process to reconstruct fluent text that adheres to a target semantic distribution. We demonstrate that our framework is naturally compatible with Differential Privacy (DP): once the semantic embeddings of the private corpus are protected, the DP guarantee carries over to the generated text by post-processing. Empirically, we show that our approach generates high-quality synthetic text with precise semantic alignment. Furthermore, we validate its effectiveness in DP synthetic text generation: across two datasets and two diffusion backbones, it outperforms Private Evolution and DP-SGD-based fine-tuning in distributional fidelity and downstream utility, while generating significantly faster than Private Evolution. We further extend the framework to Data-Specific Indistinguishability (DSI), which adds a formal certificate against data poisoning and memorization at a small utility cost.