De-Anonymization at Scale via Tournament-Style Attribution

TL;DR

This paper presents DAS, a scalable LLM-based method for authorship de-anonymization that effectively links anonymous texts to their authors in large datasets, revealing significant privacy risks.

Contribution

DAS introduces a novel, scalable LLM-driven approach with a sequential grouping and voting strategy for large-scale authorship attribution, outperforming prior methods.

Findings

DAS accurately links texts to authors in datasets of tens of thousands.

DAS outperforms previous methods in accuracy and scalability.

Experiments reveal privacy risks in anonymous platforms using DAS.

Abstract

As LLMs rapidly advance and enter real-world use, their privacy implications are increasingly important. We study an authorship de-anonymization threat: using LLMs to link anonymous documents to their authors, potentially compromising settings such as double-blind peer review. We propose De-Anonymization at Scale (DAS), a large language model-based method for attributing authorship among tens of thousands of candidate texts. DAS uses a sequential progression strategy: it randomly partitions the candidate corpus into fixed-size groups, prompts an LLM to select the text most likely written by the same author as a query text, and iteratively re-queries the surviving candidates to produce a ranked top-k list. To make this practical at scale, DAS adds a dense-retrieval prefilter to shrink the search space and a majority-voting style aggregation over multiple independent runs to improve robustness and ranking precision. Experiments on anonymized review data show DAS can recover same-author texts from pools of tens of thousands with accuracy well above chance, demonstrating a realistic privacy risk for anonymous platforms. On standard authorship benchmarks (Enron emails and blog posts), DAS also improves both accuracy and scalability over prior approaches, highlighting a new LLM-enabled de-anonymization vulnerability.