{"id":122693,"date":"2026-03-04T13:56:18","date_gmt":"2026-03-04T13:56:18","guid":{"rendered":"https:\/\/ekamu.net\/?p=122693"},"modified":"2026-03-04T13:56:18","modified_gmt":"2026-03-04T13:56:18","slug":"internette-anonimlik-bitiyor-ve-bu-yapay-zeka-sayesinde-oluyor","status":"publish","type":"post","link":"https:\/\/ekamu.net\/index.php\/2026\/03\/04\/internette-anonimlik-bitiyor-ve-bu-yapay-zeka-sayesinde-oluyor\/","title":{"rendered":"\u0130nternette anonimlik bitiyor ve bu, yapay zeka sayesinde oluyor"},"content":{"rendered":"

<\/span> B\u00fcy\u00fck dil modellerinin <\/strong>(LLM) \u00e7evrimi\u00e7i ortamlarda takma ad kullanan ki\u015fileri ger\u00e7ek kimlikleriyle e\u015fle\u015ftirebildi\u011fini <\/strong>ortaya koyan yeni bir ara\u015ft\u0131rma, internet anonimli\u011fine ili\u015fkin ezberleri bozdu. \u00c7al\u0131\u015fmaya g\u00f6re yapay zeka sistemleri, farkl\u0131 platformlardaki i\u00e7erikleri analiz ederek kullan\u0131c\u0131lar\u0131 y\u00fcksek isabet <\/strong>oran\u0131yla tespit edebiliyor. <\/figure>\n

Deneylerde geri \u00e7a\u011f\u0131r\u0131m oran\u0131 y\u00fczde 68\u2019e, do\u011fruluk oran\u0131 ise y\u00fczde 90\u2019a kadar ula\u015ft\u0131<\/strong>. Geri \u00e7a\u011f\u0131r\u0131m oran\u0131, anonim hesaplar\u0131n ne kadar\u0131n\u0131n ba\u015far\u0131yla kimlikle e\u015fle\u015ftirildi\u011fini, do\u011fruluk oran\u0131 ise yap\u0131lan e\u015fle\u015ftirmelerin ne kadar\u0131n\u0131n do\u011fru \u00e7\u0131kt\u0131\u011f\u0131n\u0131 g\u00f6steriyor.<\/p>\n

Basit bir metinden kimli\u011fe ula\u015f\u0131labiliyor<\/b><\/p>\n

<\/span> Ara\u015ft\u0131rmada kamuya a\u00e7\u0131k sosyal medya verilerinden olu\u015fturulan farkl\u0131 veri setleri test edildi. \u00d6rne\u011fin bir deneyde Hacker News \u00fczerindeki payla\u015f\u0131mlar ile LinkedIn profilleri \u00e7apraz referanslar \u00fczerinden e\u015fle\u015ftirildi. Profillerdeki a\u00e7\u0131k kimlik g\u00f6stergeleri temizlendikten sonra i\u00e7erikler LLM\u2019e verildi ve sistem kullan\u0131c\u0131lar\u0131 anlaml\u0131 do\u011fruluk oranlar\u0131yla ili\u015fkilendirebildi. <\/figure>\n
<\/span> Ge\u00e7mi\u015fte de kullan\u0131c\u0131 tercihleri ve i\u015flem kay\u0131tlar\u0131 \u00fczerinden klasik y\u00f6ntemlerle kimlik \u00e7\u0131kar\u0131m\u0131 yap\u0131labildi\u011fi daha \u00f6nce g\u00f6sterilmi\u015fti. Ancak yeni \u00e7al\u0131\u015fmada LLM tabanl\u0131 yakla\u015f\u0131m\u0131n, bu geleneksel ve daha zahmetli tekni\u011fi belirgin bi\u00e7imde geride b\u0131rakt\u0131\u011f\u0131 g\u00f6r\u00fcld\u00fc. <\/figure>\n

Ara\u015ft\u0131rmac\u0131lara g\u00f6re en kritik fark, yapay zekan\u0131n yap\u0131land\u0131r\u0131lm\u0131\u015f veri setlerine ihtiya\u00e7 duymadan <\/strong>serbest metinleri analiz edebilmesi. LLM\u2019ler internet \u00fczerinde gezinerek, i\u00e7erikleri ba\u011flamsal olarak de\u011ferlendirerek ve olas\u0131 e\u015fle\u015fmeleri ak\u0131l y\u00fcr\u00fctme s\u00fcre\u00e7leriyle test ederek kimlik tespiti yapabiliyor.<\/p>\n

Payla\u015f\u0131m artt\u0131k\u00e7a risk b\u00fcy\u00fcyor<\/b><\/p>\n

<\/span> 2024 y\u0131l\u0131na ait Reddit verileri \u00fczerinde yap\u0131lan deneyler, dijital ayak izinin b\u00fcy\u00fckl\u00fc\u011f\u00fc ile tespit edilebilirlik aras\u0131nda do\u011frudan bir ili\u015fki oldu\u011funu g\u00f6sterdi. Etkinin g\u00fcc\u00fc g\u00f6sterilmesi i\u00e7in film tart\u0131\u015fmalar\u0131n\u0131n y\u00fcr\u00fct\u00fcld\u00fc\u011f\u00fc alt forumlarda yap\u0131lan yorumlar incelendi. <\/figure>\n

Sadece tek bir film hakk\u0131nda<\/strong> payla\u015f\u0131m yapan kullan\u0131c\u0131lar\u0131n ortalama y\u00fczde 3,1\u2019i<\/strong> y\u00fczde 90 do\u011fruluk oran\u0131yla e\u015fle\u015ftirilebilirken, bu oran y\u00fczde 99 do\u011fruluk seviyesinde y\u00fczde 1,2<\/strong>\u2019ye d\u00fc\u015ft\u00fc.<\/p>\n

Be\u015f ila dokuz film<\/strong> hakk\u0131nda yorum yapanlarda tespit oran\u0131 y\u00fczde 90 do\u011frulukta y\u00fczde 8,4\u2019e<\/strong> y\u00fckseldi. Ondan fazla film<\/strong> payla\u015fan kullan\u0131c\u0131lar i\u00e7in ise y\u00fczde 90 do\u011fruluk seviyesinde kimlik tespiti oran\u0131 y\u00fczde 48,1\u2019e<\/strong> \u00e7\u0131kt\u0131.<\/p>\n

Anonimlik mi?<\/b><\/p>\n

<\/span> Teknik geli\u015fmelerin yaratt\u0131\u011f\u0131 risk, sosyal medyan\u0131n k\u00fclt\u00fcrel d\u00f6n\u00fc\u015f\u00fcm\u00fcyle birle\u015fti\u011finde tablo daha da \u00e7arp\u0131c\u0131 hale geliyor. <\/figure>\n

Ak\u0131ll\u0131 telefonlar\u0131n yayg\u0131nla\u015fmas\u0131yla birlikte g\u00f6zetim, devletlerin tekelinden \u00e7\u0131karak kitlesel bir prati\u011fe d\u00f6n\u00fc\u015ft\u00fc<\/strong>. Konserlerde \u201ckiss cam<\/strong>\u201d g\u00f6r\u00fcnt\u00fclerinden restoranlardaki s\u0131radan anlara kadar bir\u00e7ok sahne saniyeler i\u00e7inde milyonlara ula\u015fabiliyor. Viral olan i\u00e7eriklerin ard\u0131ndan kimliklerin ortaya \u00e7\u0131kar\u0131lmas\u0131 ise \u00e7o\u011fu zaman yaln\u0131zca birka\u00e7 yorum meselesi. \u0130nternet kullan\u0131c\u0131lar\u0131, video alt\u0131ndaki yorumlarda ki\u015filerin isimlerini, sosyal medya hesaplar\u0131n\u0131 ve hatta i\u015f bilgilerini payla\u015fabiliyor.<\/p>\n

Bu s\u00fcreci h\u0131zland\u0131ran unsurlardan biri de y\u00fcz tan\u0131ma teknolojileri. \u00d6rne\u011fin PimEyes<\/strong> gibi sistemler, yaln\u0131zca bir foto\u011fraftan hareketle ki\u015finin internetteki di\u011fer g\u00f6rsellerine ula\u015fmay\u0131 m\u00fcmk\u00fcn k\u0131l\u0131yor. Bu da kamusal alanda \u00e7ekilen bir g\u00f6r\u00fcnt\u00fcn\u00fcn h\u0131zla ger\u00e7ek kimlikle e\u015fle\u015ftirilmesini kolayla\u015ft\u0131r\u0131yor.<\/p>\n

Yapay zeka destekli kimlik tespiti ile viral k\u00fclt\u00fcr birle\u015fti\u011finde ortaya g\u00fc\u00e7l\u00fc bir \u201ckat\u0131l\u0131mc\u0131<\/strong> g\u00f6zetim<\/strong>\u201d d\u00fczeni \u00e7\u0131k\u0131yor. Bu d\u00fczen i\u00e7inde herhangi biri, s\u0131radan bir g\u00fcn\u00fcnde kayda al\u0131narak internetin g\u00fcndemine ta\u015f\u0131nabiliyor. Sonras\u0131nda ise \u201cdoxxing\u201d, i\u015f kayb\u0131, sosyal \u00e7evre bask\u0131s\u0131 ya da bo\u015fanma gibi ciddi sonu\u00e7lar do\u011fabiliyor.<\/p>\n

Ara\u015ft\u0131rmac\u0131lar, LLM\u2019lerin geli\u015fmeye devam etmesi halinde devletlerin \u00e7evrimi\u00e7i ele\u015ftirmenleri tespit etmek i\u00e7in bu ara\u00e7lardan yararlanabilece\u011fini, \u015firketlerin hiper hedefli reklamc\u0131l\u0131k amac\u0131yla ayr\u0131nt\u0131l\u0131 kullan\u0131c\u0131 profilleri olu\u015fturabilece\u011fini ve k\u00f6t\u00fc niyetli akt\u00f6rlerin ki\u015fiselle\u015ftirilmi\u015f sosyal m\u00fchendislik sald\u0131r\u0131lar\u0131 geli\u015ftirebilece\u011fini belirtiyor.<\/p>\n

Buraya kadar teknik ve k\u00fclt\u00fcrel boyutlar\u0131yla anonimli\u011fi tart\u0131\u015fm\u0131\u015f olsak da d\u00fcnya genelinde h\u00fck\u00fcmetler \u00e7ocuklar\u0131n sosyal medyaya eri\u015fimini s\u0131n\u0131rlamaya y\u00f6nelik ad\u0131mlar at\u0131yor ve platformlara g\u00fc\u00e7l\u00fc ya\u015f do\u011frulama sistemleri getirmeye \u00e7al\u0131\u015f\u0131yorlar. \u00d6rne\u011fin Avustralya, Aral\u0131k 2025 itibar\u0131yla 16 ya\u015f\u0131n alt\u0131ndaki \u00e7ocuklar\u0131n sosyal medya kullan\u0131m\u0131n\u0131 yasaklayan ilk \u00fclke oldu. Sosyal medya platformlar\u0131, kullan\u0131c\u0131lar\u0131n ya\u015flar\u0131n\u0131 do\u011frulamak i\u00e7in resmi kimlik, y\u00fcz veya ses tan\u0131ma gibi birden fazla y\u00f6ntem uygulamak zorunda. Bu trend \u015fimdi T\u00fcrkiye dahil her \u00fclkeye yay\u0131lm\u0131\u015f durumda.<\/p>\n\n

<\/span><\/p>\n

Kaynak :\u00a0https:\/\/www.donanimhaber.com\/internette-anonimlik-bitiyor-ve-bu-yapay-zeka-sayesinde-oluyor–202813<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

B\u00fcy\u00fck dil modellerinin (LLM) \u00e7evrimi\u00e7i ortamlarda takma ad kullanan ki\u015fileri ger\u00e7ek kimlikleriyle e\u015fle\u015ftirebildi\u011fini ortaya koyan yeni bir ara\u015ft\u0131rma, internet anonimli\u011fine ili\u015fkin ezberleri bozdu. \u00c7al\u0131\u015fmaya g\u00f6re yapay zeka sistemleri, farkl\u0131 …<\/p>\n","protected":false},"author":1,"featured_media":122694,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[7416,4518,1697,431,644],"class_list":["post-122693","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-teknoloji","tag-dogruluk","tag-kimlik","tag-sosyal-medya","tag-tek","tag-tespit"],"_links":{"self":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/122693","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/comments?post=122693"}],"version-history":[{"count":1,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/122693\/revisions"}],"predecessor-version":[{"id":122700,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/122693\/revisions\/122700"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/media\/122694"}],"wp:attachment":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/media?parent=122693"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/categories?post=122693"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/tags?post=122693"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}