[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/pol/ - Politically Incorrect


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


It looks like AI can now unmask any anonymous account on the internet. That’s according to a new study by Simon Lermen (MATS), Daniel Paleka (ETH Zurich), Joshua Swanson (ETH Zurich), Michael Aerni (ETH Zurich), Nicholas Carlini (Anthropic), and Florian Tramèr (ETH Zurich), published on arXiv.

In the paper, “Large-Scale Online Deanonymization with LLMs,” the researchers show that modern large language models (LLMs) can re-identify people behind pseudonymous online accounts at a scale and accuracy that far surpass previous techniques.

The core contribution is an automated deanonymization pipeline powered by LLMs, according to the new study. Instead of relying on structured datasets or hand-engineered features—like earlier attacks on the Netflix Prize dataset—the system works directly on raw, unstructured text.

Given posts, comments, or interview transcripts written under a pseudonym, the pipeline extracts identity-relevant signals, searches for likely matches using semantic embeddings, and then uses higher-level reasoning to verify the most promising candidates while filtering out false positives. The result is a scalable attack that mirrors—and in some cases exceeds—the effectiveness of a dedicated human investigator.

To evaluate their approach, the researchers constructed three datasets with known ground truth. The first links pseudonymous Hacker News users to real-world LinkedIn profiles, relying on cross-platform clues embedded in public text. The second matches users across movie discussion communities on Reddit. The third takes a single Reddit user’s history, splits it into two time-separated profiles, and tests whether the system can reconnect them.


https://www.zerohedge.com/markets/ai-can-now-unmask-anonymous-internet-users-new-study-finds
>>
Across all three settings, LLM-based methods dramatically outperformed classical baselines, which often achieved near-zero recall.

The headline numbers are striking. In some experiments, the system achieved up to 68% recall at 90% precision—meaning it correctly identified a substantial portion of targets while keeping false accusations low. Even when matching temporally split Reddit accounts separated by a year, performance remained strong. In contrast, traditional non-LLM approaches struggled to produce meaningful matches. The findings suggest that advances in reasoning and representation learning have transformed deanonymization from a niche, data-hungry attack into a broadly applicable capability.

The study says that a key concern is that the attack pipeline is composed of individually benign steps: summarizing text, generating embeddings, ranking candidates, and reasoning over matches. No single component appears inherently malicious, making it difficult to detect or restrict through conventional safeguards. Moreover, the study finds that increasing model reasoning effort improves deanonymization performance, implying that as frontier models become more capable, the attack may become even more effective by default.

The broader implication is that “practical obscurity”—the idea that scattered, pseudonymous posts are safe because linking them is too labor-intensive—may no longer hold.

Persistent usernames, writing style, niche interests, and cross-platform references can collectively act as a fingerprint. The authors conclude that threat models for online privacy need to be reconsidered in light of LLM capabilities. While not every account can be unmasked, and performance varies by context, the study makes clear that the technical barrier to large-scale deanonymization has fallen dramatically.
>>
>>529357066
>Anonymous account

Wonderful oxymoron there
>>
>>529357066
jokes on them I never type the same shit. Ai is a hoax.
>>
File: 1771503598342240.jpg (137 KB, 1200x1200)
137 KB
137 KB JPG
>posting on social media accounts
>>
>>529357066
>according to the new study
funded by Ai. It's like rockefeller pharma funding it's own studies to come to a predecided conclusions. Go fuck yourselves fat bastards. Ai is designed by a bunch of stupid jews that cannot create or build anything of value in the first place. Thanks for the laughs
>>
>>529357066
I purposely change my typing style on here and it's the only place on the internet i post, suck my dick and good luck fags
>>
File: 1683067633274448.jpg (115 KB, 922x851)
115 KB
115 KB JPG
>>529357532
Good luck anon
>>
>>529357267
The research is publicly funded by the Swiss government, no money from AI companies https://arxiv.org/pdf/2602.16800
>>
>>529357644
>The research is publicly funded by the Swiss government
Swiss mountain jews funding this research? wow I must trust the Science!
>>
>>529357836
ETH Zurich is one of the top 20 research universities in the world and the names of the resarchers aren't even jewish dumbass
>>
>>529357532
Same, I'm paranoid as hell so I randomly switch things up and change my E-mail and shit as if I am being hunted. Not that I know of anyone doing so, but just in case, ya know?
>>
>>529357066
Grok, who is OP and how many cocks has he sucked?
>>
>>529357885
>Joshua
HAHAHAHAHAHAH sure mountain jew.
>>
You realise these AI companies release these bullshit marketing stories as a way to scare the goyim into submission right?
>don’t be le antisemitic goy, we are in your walls, we know what you jack off to

Lol fuck off kikes. Jews can suck my dick and tongue my anus for good measure.
>>
>>529357885
''Joshua is a prominent Hebrew name derived from Yehoshua (יְהוֹשֻׁעַ), meaning "Yahweh is salvation" or "God is deliverance"
HAHAHAHAHAHA
>>
>>529357984
>>529358040
How dumb are you man? The last name is 'Swanson', one of the most white anglo-saxon names that exists
>>
>>529357885

sweet summer child

tribe of dan are identity thieves
>>
Blackmail me for my bitcoin

I’ll give you my hot crush’s face, her first name, and my small dick. If you find her I’ll give you her number. If you send me a bikini pic of her within 10 minutes I’ll give you my face too.

For each picture you find of us, I’ll give you a letter of her last name.

Blackmail me for whatever you want.

Kik me: amandascc
>>
>>529357066
Joke’s on them, I was only pretending to be retarded
>>
>>529358086
jewish parasites steal white family names to blend in. More news at 6.
>>
>>529357066
Well I would like to just say for the record FED NIGGERS TONGUE MY ANUS.
>>
>>529357066
It's a good thing that I havent ever told anybody I am a gay 53 year old episcopalian man with three cats before...! Or else the AI would have figured me out by now.
>>
File: 855a2321ba9.jpg (36 KB, 333x500)
36 KB
36 KB JPG
>>529357066
this is why I randomly change my typing signatures, filenames, and personas
>>
>>529357156
fippity bippity
>>
>>529357984
Hate to tell you this but Joshua/Josh was the most popular name in England for fucking ages.
Josh for boys and Bethany for girls
I even know couples that had one of each.
People are fucking sheep with kids names
>>
>>529357066
>The first links pseudonymous Hacker News users to real-world LinkedIn profiles, relying on cross-platform clues embedded in public text.

I'm cooked.
>>
>>529357070
This is actually not super impressive. If you're in the bottom 70th percentiles of "I post PII about myself on the internet", an AI might accuse you of being someone, and it will be wrong 10% of the time.

The typical redditor isnt conspiratorial enough, but most of /pol/ I assume would be, to never disclose such details even under an anon account. At best, your industry, marital status, state... not your company or job title, not your exact age and child/pet count, not your city or reams of posts in r/(insert your town) that make it obvious you are local.

That said, an AI with access to telecom data and the reddit (or, say, x.com or 4chan) db, with IP addresses and browser fingerprints? We're 99.9% identifiable by an AI, no question. I guess we will be outed by a government or sophisticated actor, while retarded x/reddit/etc. Anons will get outed from public info slightly earlier.
>>
File: 1749228510527095.gif (342 KB, 487x603)
342 KB
342 KB GIF
>>529357066
tfw claude orders pizza to my house after cussing it for paywall prompting
>>
>>529357066
Just be horny on main you fucking faggot
No one even cares
It's no longer "I don't have anything to hide!"
Now it's "Do you really want to look, sicko? I dare you."
>>
File: 1770708256902692.gif (1.15 MB, 358x200)
1.15 MB
1.15 MB GIF
>accounts
>>
>>529358854
if someone walks in on you jacking off and you stop, you got caught jacking off
if you keep beating instead, now he's the weirdo who watched a guy wax his weasel
>>
>>529357066
>we used a bunch of words to pretend idiolect is new
whoa that's amazing
the future is here
>>
>>529357066
little did they know i consciously use improper grammar online along with similes and vocabulary I never use prfoessionally
>>
>>529357532
>not swapping your style between posts/threads
Never got called a samefag



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.