Zero-Shot Privacy-Aware Text Rewriting via Iterative Tree Search

2025-09-25

AI Summary

To address privacy leakage risks of sensitive user information in large language model (LLM) applications deployed on cloud services, this paper proposes a zero-shot, training-data-free iterative tree search framework for text rewriting. The method dynamically identifies and replaces or removes privacy-bearing segments via structured tree search, integrating sentence-level rewriting operations with reward-model-guided progressive optimization to achieve privacy-utility trade-offs without labeled data. Its core innovations are: (i) the first integration of zero-shot learning with tree search for privacy-aware text rewriting; and (ii) a novel joint evaluation mechanism that jointly assesses privacy removal, semantic consistency, and linguistic naturalness. Extensive experiments across multiple privacy-sensitive benchmarks demonstrate that our approach significantly outperforms existing baselines in privacy elimination rate, semantic fidelity, and fluency—effectively balancing stringent privacy protection with high textual utility.

Technology Category

• Machine Learning: Privacy
• Natural Language Processing: (Large) Language Models
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

The increasing adoption of large language models (LLMs) in cloud-based services has raised significant privacy concerns, as user inputs may inadvertently expose sensitive information. Existing text anonymization and de-identification techniques, such as rule-based redaction and scrubbing, often struggle to balance privacy preservation with text naturalness and utility. In this work, we propose a zero-shot, tree-search-based iterative sentence rewriting algorithm that systematically obfuscates or deletes private information while preserving coherence, relevance, and naturalness. Our method incrementally rewrites privacy-sensitive segments through a structured search guided by a reward model, enabling dynamic exploration of the rewriting space. Experiments on privacy-sensitive datasets show that our approach significantly outperforms existing baselines, achieving a superior balance between privacy protection and utility preservation.

Problem

• Protecting sensitive information in user inputs to cloud LLMs
• Balancing privacy preservation with text naturalness and utility
• Systematically obfuscating private data while maintaining coherence

Innovation

• Tree-search-based iterative sentence rewriting algorithm
• Zero-shot privacy-aware text rewriting method
• Dynamic exploration guided by reward model