The internet is rife with anonymous accounts as users adopt pseudonyms, sometimes for genuine reasons like speaking freely, and other times for nefarious ones. But this era of online privacy could be coming to a close. In a study available on the arXiv preprint server, researchers demonstrate that large language models (LLMs) can identify the people behind these accounts at scale.
For years, there has been an assumption that if deanonymization were theoretically possible, it would be too time-consuming, difficult, and expensive for people to perform. But the study authors had a hunch that LLMs had become powerful enough to break online invisibility.
To test whether this was the case, the team designed an automated framework to replicate a human investigator's decision-making process.
To read more, click here.