Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Large-Scale Online Deanonymization with LLMs

simonlermen.substack.com

Post reply on HN
Print view View original

31–40 of 258 posts

Re: Large-Scale Online Deanonymization with LLMs

#35

Post by famouswaffles »

ranger_danger wrote: post #23
DalasNoin wrote: post #4

Earlier quoted context omitted.

To be clear, we are making a clear concession here that the people weren't truly anonymous. But we did use an LLM to remove any identifying information from HN making them quasi-anonymous, this is more described in the appendix Table 2. We do also make a more real world like test in section 2. There we use the anthropic interviewer dataset which Anthropic redacted, from the redacted interviews our agent identified 9/…

But you also relied on people giving away too much personal information about themselves... which won't always be the case.

Over a large enough timeframe (often a couple years at most), almost everyone online gives too much information about themselves. A seemingly innocuous statement can pin you to an exact city and so on.

Re: Large-Scale Online Deanonymization with LLMs

#36

Post by dang »

DalasNoin wrote: post #4
squeefers wrote: post #2

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

To be clear, we are making a clear concession here that the people weren't truly anonymous. But we did use an LLM to remove any identifying information from HN making them quasi-anonymous, this is more described in the appendix Table 2. We do also make a more real world like test in section 2. There we use the anthropic interviewer dataset which Anthropic redacted, from the redacted interviews our agent identified 9/…

Thanks for that link! I'll put in the top text.

Edit: actually I've re-upped your submission of that link and moved the links to the paper to the toptext instead. Hopefully this will ground the discussion more in the actual study.

Re: Large-Scale Online Deanonymization with LLMs

#37

Post by ceejayoz »

kseniamorph wrote: post #30

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.

> Most adversaries who would want to deanonymize people at scale (governments, corporations) already have access to far more direct methods.

Easier methods probably means more adversaries.

Re: Large-Scale Online Deanonymization with LLMs

#38

Post by yomismoaqui »

I did something like this passing some of my comments here and then prompted Gemini to identify my native language by reading my not-so-good english.

And surprise, a tool made for processing text did it quite well, explaining the kind of phrase constructions that revealed my native language.

So maybe this is a plus for passing any text published on the internet through a slopifier for anonymization?

EDIT: deanonymization -> anonymization

Re: Large-Scale Online Deanonymization with LLMs

#39

Post by DalasNoin »

Zigurd wrote: post #25

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.

We do advocate for stricter controls on data access on social platforms because of this. There is a bit of an unfortunate trade-off, but I think allowing mass-scraping or downloads of data from social sites can be misused in increasingly more ways.
Post reply on HN