Large-scale online deanonymization with LLMs

TL;DR

This paper demonstrates that large language models can effectively deanonymize pseudonymous online profiles across platforms by extracting features, searching for matches, and verifying identities, surpassing classical methods significantly.

Contribution

It introduces a scalable LLM-based approach for deanonymization directly on raw user content, outperforming traditional structured-data methods in multiple online settings.

Findings

LLMs achieve up to 68% recall at 90% precision in deanonymization tasks.

The approach outperforms classical baselines by a large margin.

Online privacy protections are significantly weakened by LLM-based deanonymization.

Abstract

We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator. We then design attacks for the closed-world setting. Given two databases of pseudonymous individuals, each containing unstructured text written by or about that individual, we implement a scalable attack pipeline that uses LLMs to: (1) extract identity-relevant features, (2) search for candidate matches via semantic embeddings, and (3) reason over top candidates to verify matches and reduce false positives. Compared to classical deanonymization work (e.g., on the Netflix prize) that required structured data, our approach works directly on raw user content across arbitrary platforms. We construct three datasets with known ground-truth data to evaluate our attacks. The first links Hacker News to LinkedIn profiles, using cross-platform references that appear in the profiles. Our second dataset matches users across Reddit movie discussion communities; and the third splits a single user's Reddit history in time to create two pseudonymous profiles to be matched. In each setting, LLM-based methods substantially outperform classical baselines, achieving up to 68% recall at 90% precision compared to near 0% for the best non-LLM method. Our results show that the practical obscurity protecting pseudonymous users online no longer holds and that threat models for online privacy need to be reconsidered.