CYBERSECURITYTRACKER
TRACKING6,506 stories in this site build1,309 vulnerability news stories in this site build
Permanent story citation

Risky Bulletin: LLMs can deanonymize internet users based on their past comments

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 230

As cited

Copy frozen at (site build).

ai security

Risky Bulletin: LLMs can deanonymize internet users based on their past comments

Researchers have demonstrated that large language models can deanonymize internet users by analyzing past comments and digital activity, even when individuals use different pseudonyms across platforms. The technique creates user profiles based on linguistic patterns, shared vocabulary, and contextual clues like location and interests to link real identities to anonymous or pseudonymous accounts.

Why it matters: Security practitioners and privacy-conscious users should understand this deanonymization capability as a potential threat to pseudonymous security research, whistleblowing, and other sensitive online activities.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary