Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces AURA, a novel LLM-powered mask-reconstruct framework designed to enhance text anonymization while preserving utility in the face of agentic re-identification threats posed by web search. By decoupling privacy localization from utility retention, AURA employs adversarial checks to optimize the balance between privacy and the analytic value of the text. Evaluation on real-user interview transcripts demonstrates that AURA significantly improves the privacy-utility frontier, effectively resisting re-identification while maintaining contextual integrity.
AURA redefines the privacy-utility balance in text anonymization, achieving stronger resistance to agentic re-identification without sacrificing analytic value.
Agentic LLMs with web search change the threat model for text anonymization: weak contextual cues can become cross-referenceable evidence for re-identification, yet those same details also carry downstream analytic value of the text. Existing defenses either remove explicit identifiers, perturb text for formal privacy, or test rewritten text against non-web inference models, leaving underexplored the operating region between resistance to agentic web-search re-identification and utility retention. We introduce AURA (\textbf{A}nonymization with \textbf{U}tility-\textbf{R}etention \textbf{A}daptation), an LLM-powered \textit{mask-reconstruct} framework that decouples privacy localization from utility-preserving reconstruction and selects candidates with adversarial privacy and utility-retention checks. We evaluate AURA on real-user interview transcripts using re-identification attacks carried out by web-search agents, along with a utility evaluation based on interviewee-profile facts, codebook facts, and the joint contextual utility grid. Our results show that AURA improves the privacy-utility frontier by using adaptive privacy scope to strengthen resistance to agentic re-identification and using a mask-reconstruct anonymization method to better preserve contextual utility under fixed privacy scope.