LLM-Based System Achieves 68% Recall at 90% Precision for Online User Deanonymization

Researchers demonstrate that large language models can effectively deanonymize online users by analyzing their writing style and content across platforms. Their system matches 68% of true user pairs with 90% precision, significantly outperforming traditional methods. LLMs Can Now Deanonymize Online Users with 90% Precision by Analyzing Writing Patterns A new research paper titled "Large-scale online deanonymization with LLMs" demonstrates that anonymous usernames provide diminishing protection against modern AI systems. The study shows that large language models can piece together a person's public trail across different platforms by analyzing their writing style and content, achieving 68% recall at 90% precision—meaning 9 out of 10 matches are correct. What the Researchers Built The research team developed a three-stage LLM-based pipeline for linking anonymous user accounts across different online platforms. Unlike traditional methods that rely on exact string matching or simple metad