Skip to main content
Search

# AURA

1 related articles

Subscribe to updates
  1. arXiv

    AURA combines mask-reconstruct anonymization with adversarial screening to reach the lowest agentic re-identification counts among non-DP methods and recover 6.4 percentage points more contextual utility than a prior LLM anonymizer at comparable privacy

    The work introduces AURA (Anonymization with Utility-Retention Adaptation), an LLM-powered mask-reconstruct framework that decouples privacy localization from utility-preserving reconstruction and selects candidates through adversarial privacy and utility-retention checks; evaluated on real-user interview transcripts with re-identification attacks carried out by web-search agents, adaptive-scope AURA yields the lowest agentic re-identification counts under each of three attacker models among non-DP methods, and at matched scope and backbone it retains more contextual utility than the prior LLM anonymizer (+6.4 pp unit-grid recovery) at comparable privacy.

You’re all caught up