D_a_X@feddit.org to Netzkultur / Netzpolitik@feddit.org · 2 months agoLLMs can unmask pseudonymous users at scale with surprising accuracy - Ars Technicaarstechnica.comexternal-linkmessage-square2linkfedilinkarrow-up128arrow-down11file-textcross-posted to: privacy@programming.devfuck_ai@lemmy.worldprivacy@programming.devprivacy@lemmy.dbzer0.comprivacy@lemmy.worldtechnology@lemmy.ziptechnologie@jlai.luhackernews@lemmy.bestiver.secybersecurity@sh.itjust.works
arrow-up127arrow-down1external-linkLLMs can unmask pseudonymous users at scale with surprising accuracy - Ars Technicaarstechnica.comD_a_X@feddit.org to Netzkultur / Netzpolitik@feddit.org · 2 months agomessage-square2linkfedilinkfile-textcross-posted to: privacy@programming.devfuck_ai@lemmy.worldprivacy@programming.devprivacy@lemmy.dbzer0.comprivacy@lemmy.worldtechnology@lemmy.ziptechnologie@jlai.luhackernews@lemmy.bestiver.secybersecurity@sh.itjust.works
minus-squareKairos@lemmy.todaylinkfedilinkarrow-up6·2 months agoHi. Resident skeptic here. How did they make the data? There’s effectively zero way to run this on real accounts (no way to verify correctness) so I’m thinking that the researchers made accounts that aren’t realistic.
minus-squareGeneral_Effort@lemmy.worldlinkfedilinkarrow-up1·2 months agohttps://arxiv.org/abs/2602.16800
Hi. Resident skeptic here. How did they make the data? There’s effectively zero way to run this on real accounts (no way to verify correctness) so I’m thinking that the researchers made accounts that aren’t realistic.
https://arxiv.org/abs/2602.16800