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 prior deanonymization work (e.g., on the Netflix prize) that required structured data or manual feature engineering, our approach works directly on raw user content across arbitrary platforms.
More: And right on schedule: there goes pseudonymity on the Internet.