Forgot your password?
Close
typodupeerror
AI The Internet Privacy

AI Allows Hackers To Identify Anonymous Social Media Accounts, Study Finds (theguardian.com) 54

Posted by BeauHD from the heads-up dept.
An anonymous reader quotes a report from the Guardian: AI has made it vastly easier for malicious hackers to identify anonymous social media accounts, a new study has warned. In most test scenarios, large language models (LLMs) -- the technology behind platforms such as ChatGPT -- successfully matched anonymous online users with their actual identities on other platforms, based on the information they posted. The AI researchers Simon Lermen and Daniel Paleka said LLMs make it cost effective to perform sophisticated privacy attacks, forcing a "fundamental reassessment of what can be considered private online".

In their experiment, the researchers fed anonymous accounts into an AI, and got it to scrape all the information it could. They gave a hypothetical example of a user talking about struggling at school, and walking their dog Biscuit through a "Dolores park." In that hypothetical case, the AI then searched elsewhere for those details and matched @anon_user42 to the known identity with a high degree of confidence. While this example was fictional, the paper's authors highlighted scenarios in which governments use AI to surveil dissidents and activists posting anonymously, or hackers are able to launch "highly personalized" scams.
This discussion has been archived. No new comments can be posted.

AI Allows Hackers To Identify Anonymous Social Media Accounts, Study Finds More

AI Allows Hackers To Identify Anonymous Social Media Accounts, Study Finds

Comments Filter:

  • a day in the park (Score:4, Interesting)

    by bugs2squash ( 1132591 ) on Monday March 09, 2026 @12:04PM (#66031332)
    I was walking my dog, Felicity, through Hegsketh park yesterday when I came across an oak tree, which, as you know, is an all consuming personal interest of mine
    • It's not that surprising using hyper specific terms across multiple accounts leads to linking them. Who could have seen that coming. Pretty sure we were doing this before LLMs anyway. Status quo is to just lie about everything and you'll keep being anon.

      • Re: (Score:2)

        by Tyr07 ( 8900565 )

        Is this even really an LLM thing?

        If you searched an account for specific phrases, places or things and found major events or issues that are in it, and google search it, which is all the LLM is basically doing, you'll probably get a hit on another social platform where they posted that exact specific thing?

        What is this? "People googling information found that AI could also google information and get similar results"

    • Re: (Score:1)

      by sinij ( 911942 )
      When I was walking in the park, I noticed that you were not picking up after your dog, Felicity. When I approached and mentioned that you forgot to pick after you dog in traditional Kiswahili, you just rudely replied and walked away. How inconsiderate of you.

  • If their personal info is not online... (Score:3, Insightful)

    by Anonymous Coward on Monday March 09, 2026 @12:14PM (#66031350)

    This only works if they actually have legitimate profiles elsewhere. I guess this means people who are truly anonymous can and will remain anonymous. It's only fools who actually post real information online.

  • but if you are spouting out localizing information and trying to be anonymous, then you are engaging in a futile effort. AI just makes it faster other than having a team of people running through information. If you are a savvy anonymous user, then you don't write localizing information in the first place or information that can tie the accounts together.

    • It won't need all that sport of information. It'll be able to recognize patterns in the usage of the words themselves.

      • As long as you aren't using any locally identifying language or specific slang then I doubt you can tell one person definitely from another via word choice alone. At least not within a sea of multiple accounts. If you were given two pieces of text and had to surmise if they were the same person sure but one to millions without a specific phrase that's unique to the author I'd be surprised.

        • As long as you aren't using any locally identifying language or specific slang then I doubt you can tell one person definitely from another via word choice alone. At least not within a sea of multiple accounts. If you were given two pieces of text and had to surmise if they were the same person sure but one to millions without a specific phrase that's unique to the author I'd be surprised.

          People do not create whole sentences de novo. We each have tendencies to re-use phrases and constructs, as well as tendencies for how we type sentences -- including this one which contains multiple clauses, commas, an em dash, a hyphenated "re-use" instead of "reuse", I wrote "de novo" instead of ex nihilo or ab chao or out of nowhere or from scratch, and so on.

          and so on
          and whatnot
          on and on
          yadda yadda
          etc.
          etc
          etcetera
          &c

          There are lots of possible variations for every choice of word and punctuation; enough

          • And companies like Google already have a decade plus head start. They've had access to our e-mails, etc. (assuming use of g-mail) for a long time, and somewhat recently decided they should have access to our text messages as well. I have no doubt that they could have their AI pick people out of a sea of people based on the learned patterns of communication. Reddit having ai loosed on it is a similar trove. I'm sure people's alt accounts in various places will quickly be noted in the background (if they have

      • Yes, but that isn't what the researchers tested, they gave specific information in their test that can link the user to the words.

  • It would be a hell of a lot more helpful.. (Score:5, Interesting)

    by geekmux ( 1040042 ) on Monday March 09, 2026 @12:30PM (#66031378)

    ..if AI helped identify the AI slop-infested dogshit acting as "anonymous" to enrage a partisan populous.

    Let's see how good it is at telling on itself. For clickbait and spams sake.

  • People thought it was to help identify military targets. Now we really know it is to identify anyone. Hence everyone is a military target. Good to know.

    • It's pretty fucking bad at identifying military targets too - Claude managed to direct missiles to kill the people Trump was hoping would take over Iran.

  • Anonymous accounts are easy to identify (Score:3, Funny)

    by greytree ( 7124971 ) on Monday March 09, 2026 @12:33PM (#66031384)
    Anonymous accounts are easy to identify - they don't have a real name on them.

    Oh, you meant the accounts were de-anonymized and the users identified ?

    Then write that.

    • Oh, you meant the accounts were de-anonymized and the users identified ? Then write that.

      The summary literally says "successfully matched anonymous online users with their actual identities on other platforms." You have poor social skills and your reading comprehension sucks.

  • I learned this lesson 27 years ago (Score:4, Interesting)

    by Tablizer ( 95088 ) on Monday March 09, 2026 @12:36PM (#66031392) Journal

    as a troll used search engines to cross-reference various clues I had inadvertently left behind on the internet to dox me. It only required determination, not rocket science. Bots have all the time in the world to clue-hop.

  • Literally no one thinks they can post their address, dog's name and car type on Facebook, post their dog's name, city and car type on Slashdot, and remain anonymous. And you SHOULD be posting lies about yourself anyway, just in case. Post a snippet about being too cold in late summer. Post about your pet parakeet when you have a cat. Post about your brother when you're an only child.

  • Lying to yourself is the biggest danger for trying to stay Anonymous. With enough patterns to recognize, the idea that one can hide is a delusional take.

    The only way to win, is to run EVERYTHING you post through an AI that changes the tone and words used in all your online activity. But even then that may itself be a lie.

  • This is why I use Ai to rephrase my comments so that Ai can identify itself as the poster

  • Like, Facebook did this nearly 20 years ago now when it would suggest new people to follow. JFC it was creepy when a throw-away fake account would get suggestions on who to follow and it would be both a friend's secret account to hide from her stalker, and also her abuser/ex/stalker.

    Big tech is a big problem for victims of domestic violence.

    But hey, it's all worth it for the optimal personalized mattress sales and increase in shareholder value.

  • ... on other platforms. Let me think ... nope. Don't have any. Lots of phoney IDs. There is no law against me photoshopping my face on a fake "government" ID and presenting it to my laptop camera for social media accounts that require it. So long as I don't present it where an ID requirement has a basis in law. And I don't scan/photograph those. I walk them into the bank.

    Now signing off,
    Charles Ulysses Farley

  • You don't need AI to do this. This is simple big data assembling profiles of people who are sloppy enough to post identifying information. Using AI to do it because you're too incompetent or lazy to do it otherwise isn't newsworthy, and it's not yet another "AI story."

  • The word you're looking for is statistics. And here AI models (LLM or not) really show that they are good at fitting data.

    But please don't act too surprised. I bet the three letter agencies are doing that since years. Alone the time correlation between posts, if you have post data of different platforms on a large scale, will probably give away many users. It's not only constraints (when you sleep, you post on none of these platforms, you can type only on one at a time, etc.), but also humans following patt

  • Nothing is private anymore and you might as well get used to it.

    Practically *everything* you do or have done has left a "paper trail"...and now it's safely stored in gov-controlled servers forever and ever where it can be retrieved at any time to punish you.

    Just try and buy something without leaving a trail. Even with cash you have to go somewhere to get it, and the chances of you going from your home to a store and back without being seen and recorded is roughly a billion to one.

    Add in LPRs, FLOCK, video

    • Re: (Score:1)

      by sinij ( 911942 )

      Nothing is private anymore and you might as well get used to it.

      Nonsense. It is only if you do not value your privacy and go for convenience. De-Google your Android. Don't install social media apps. Secure your browser by installing decent ad-blocker. It is not that difficult to not be mostly not tracked.

      • Nonsense. It is only if you do not value your privacy and go for convenience. De-Google your Android. Don't install social media apps. Secure your browser by installing decent ad-blocker. It is not that difficult to not be mostly not tracked.

        Really? I dare you to try to travel 10 miles from your home and not be recorded or tracked. Try it. Drive, bike, walk- it doesn't matter, you'll be on camera, probably many cameras. Covering your face won't really help, in fact it may even make you more distinctive.

        Use your car? Tracked.
        Take an Uber? Tracked.
        Walking to your destination? Tracked.
        Have a friend take you? Tracked (their car, not yours!)
        Anonymous cab ride? Tracked.
        Public transportation? Definitely tracked, lol.

        Also, don't forget to leave your po

        • Re: (Score:1)

          by sinij ( 911942 )
          Recording at most would allow someone to establish time & place. It will not allow them to know my political opinions, to know them my financial situation, etc. Giving up on your privacy because of THAT is really foolish.

          My car does not have any trackers. I disabled them all on hardware level. License plate readers are thankfully rare where I live and none on my route (that I know of).

          I don't use Uber exactly because their app leaks data and they violate your privacy. Calling a regular cab and payin

          • Recording at most would allow someone to establish time & place. It will not allow them to know my political opinions, to know them my financial situation, etc. Giving up on your privacy because of THAT is really foolish.

            Really?

            Dude, you'd be surprised at what can be discerned, discovered, and determined just by tracking things like your travel and purchases, the cell phone IDs you're in proximity to, etc etc. With enough data it's not hard at all to "connect" groups of people and cross-index what's known about them.

            All of the measures you described are easy to uncover and obtain data from. You call a cab? There's a record of where it started and stopped. If you carried a phone (not just your phone, any phone), there's a re

  • ...you NEVER use your real name on 'the internets'.

    Your bank might need to know who you really are, nobody else.

    • Unfortunately, "Age verification" means identity authentication: Soon, every account will be linked to a real name, DoB and address.

      That will happen because age verification laws means corporations have the power to make subscribers into un-persons: That won't matter greatly for Twitter, BlueSky, Facebook, Slashdot, etc. But when Google/Apple/Microsoft, Ebay, computer-run tournaments (Eg. Bandai One Piece), do it, the result is damage to both one's meat-space life and one's anonymity.

  • There was a talk about this already last year at BSides TLV:
    https://bsidestlv.com/agenda/8... [bsidestlv.com]

  • There's lots of herp-derping here from people absolutely convinced they have not left enough information to be identified.

    But that only affects being *correctly* identified.

    Hiding your information doesn't help when an AI hallucinates your membership in the antifa terrorist organization. Note that antifa is not an organization, therefor you can't be a member. However, it has already been shown that mere logic and reality won't stop a certain type of people from turning that bogus claim into a real threat.

Slashdot Top Deals

I don't have any use for bodyguards, but I do have a specific use for two highly trained certified public accountants. -- Elvis Presley

Close