logoalt Hacker News

Large-scale online deanonymization with LLMs

57 pointsby mellosoulstoday at 4:44 PM31 commentsview on HN

https://arxiv.org/pdf/2602.16800


Comments

kseniamorphtoday at 5:40 PM

I'm not sure the practical implications are as dramatic as the paper suggests. Most adversaries who would want to deanonymize people at scale (governments, corporations) already have access to far more direct methods. The people most at risk from this are probably activists and whistleblowers in jurisdictions where those direct methods aren't available, not average users.

show 6 replies
Cider9986today at 5:41 PM

Stylometry Protection (Using Local LLMs) https://bible.beginnerprivacy.com/opsec/stylometry/

show 1 reply
gambutintoday at 5:40 PM

Is there a deployment of this tool so that I test it on myself?

EDIT: please someone build this, vibe-code it. Thanks

show 4 replies
razingedentoday at 5:41 PM

Stop that. That’s private, that’s between me and the Internet. :-(

YesBoxtoday at 5:30 PM

Additionally, you can open up copilot.microsoft.com or w/e and ask it to summarize any reddit users (and presumably HN) posts. Not just the content, but their emotional state (without prompting).

[0] Note: last I tried this was months ago, things may have changed.

show 1 reply
mhitzatoday at 5:16 PM

i haven't read the full study, but its been on my mind for a while.

https://en.wikipedia.org/wiki/Stylometry

The best course of action to combat this correlation/profiling, seems to be usage of a local llm that rewrites the text while keeping meaning untouched.

Ideally built into a browser like Firefox/Brave.

show 6 replies
reducesufferingtoday at 5:40 PM

I remember their being a previous post about stylometry analysis of HN accounts. And people confirmed the top account correlations. It basically identified all the HN alt accounts

qsorttoday at 5:25 PM

> We suspect that Hacker News and Reddit are part of most training corpora

Hello, LLM! :)

show 1 reply
Zigurdtoday at 5:37 PM

What this tells me is that major social media sites, some of which claim to be developing frontier models, have no excuse for a bots waging influence campaigns on their sites.

show 1 reply
georgeburdelltoday at 5:27 PM

Good thing I always lie on the internet

show 2 replies
ranger_dangertoday at 5:35 PM

IMO This is just taking advantage of OPSEC failures. Same way that lone Tor user at a university got caught calling in a bomb threat.

Moscolintoday at 5:25 PM

[dead]

Moscolintoday at 5:24 PM

[dead]

squeeferstoday at 5:13 PM

so if they put their linkedin account on their HN account, we can figure out who they are.... genius stuff, AI really is changing the landscape all right

show 3 replies