Distributed method protects privacy while creating synthetic text data
Distributed and Private Textual Data Synthesis from Embeddings
Cryptography and Security
Summary
Sharing sensitive text data from many people usually needs a trusted middleman, which can risk privacy. This work designs a new way to create synthetic text data that protects privacy without needing a trusted person to see the original texts. The approach summarizes user data as private text embeddings, then generates new text from them without repeated user involvement. Their method keeps rare user info safe and achieves similar accuracy to older approaches requiring central control.
What this means in practice
- •For data privacy engineers: Generate synthetic textual data from decentralized user inputs without exposing raw texts or relying on a trusted central server.
- •For enterprise data teams: Produce differentially private synthetic text data to support analysis and model training while preventing leakage of sensitive user information.
Authors
Ergute Bao, Hongyan Chang, Ali Shahin Shamsabadi, Ting Yu, Xiaokui Xiao
Abstract
We revisit differentially private (DP) text synthesis in the realistic setting of distributed users, where privacy concerns preclude a trusted curator with access to raw user texts. Existing DP text synthesis pipelines are designed for a trusted, centralized curator and often cannot be deployed in distributed settings due to unrealistic trust and access assumptions; when adapted naively, they require repeated, tightly synchronized user participation and incur significant overhead. To address this gap, we propose a DP--cryptography co-design for textual data synthesis that requires no trusted curator and requires only lightweight user participation. Our approach has two optimized components. First, we design a distributed-friendly DP synthesis algorithm that releases a one-time DP summary in an embedding space: it identifies frequent semantic regions and releases their DP centroids, enabling training-free, non-iterative offline text synthesis. We further introduce semantic support protection, which ensures the released summary avoids semantic neighborhoods of infrequent texts, reducing the risk of exposing rare user data. Second, we develop a custom secure protocol that implements this algorithm over distributed user data, enforcing end-to-end DP guarantees without requiring a trusted curator. On four benchmarks, we achieve utility comparable to the state-of-the-art centralized DP synthesis method.