kobuchwepheshe be-deepfake kudale izinselelo ezinkulu ekuvikelekeni kwedijithali. E-Brazil, lolu hlobo lokukhwabanisa lusabalala ngokushesha: ngo-Okthoba 2024, Amaphoyisa Omphakathi Wesifunda Sikahulumeni wethula Umsebenzi othi "DeGenerative AI," ngenhloso yokuhlakaza iqembu lezigelekeqe eligxile ekugedeni ama-akhawunti asebhange lisebenzisa izicelo zobuhlakani bokwenziwa.
Iqembu eliphenywayo lenze imizamo engaphezu kuka-550 yokugebenga ama-akhawunti asebhange abanikazi bama-akhawunti asebhange edijithali ngokuhlaselwa okuhleliwe, ukusetshenziswa kwedatha yomuntu wesithathu, kanye nobuchwepheshe be-deepfake. Besebenzisa le ndlela, bakwazile ukukhiqiza kabusha izithombe zabanikazi be-akhawunti ukuze baqinisekise izinqubo zokuvula i-akhawunti futhi banike amandla amadivayisi amasha. Iqembu lezigelekeqe likwazile ukukhipha imali elinganiselwa ku-R110,000,000 ngama-akhawunti abantu ngabanye kanye nezinkampani, emisebenzini ephakamisa ukushushumbiswa kwemali—umonakalo awubanga mubi kakhulu ngenxa yocwaningomabhuku lwamabhange okuvimbela ukukhwabanisa, olukwazile ukumisa ingxenye enkulu yokukhwabanisa.
Ubuchwepheshe be-Deepfake buguquka njalo - futhi kungenzeka bukhule nakakhulu: ngokocwaningo olwenziwa yi-Deloitte, kungenzeka ukuthola isofthiwe yokukhwabanisa kuwebhu ejulile ngezintengo ezisuka ku-US$20 ukuya ezinkulungwaneni zamadola, okubonisa amandla omnotho wokukhwabanisa emhlabeni wonke, igama elisetshenziswa i-Javelin Strategy & Research ukubhekisela kulokhu kukhula kwezenzo zobugebengu ezenziwa emhlabeni wonke, kuhlanganise nokukhwabanisa okuhlukahlukene.
Ngokombiko we-Financial Fraud Report by idwall, ukukhwabanisa okuyinkimbinkimbi kakhulu kukhuphuke ngo-16% uma kuqhathaniswa nekota yokuqala ka-2023 nekota yokuqala ka-2024. Kodwa uma sikhuluma ngobunkimbinkimbi obuphezulu, yiziphi izinhlobo zokukhwabanisa okufanele izinkampani ziziqaphele?
Kunezinhlobo ezimbili ezivame kakhulu: ukudalwa kwabasebenzisi kanye nemibhalo enedatha yokwenziwa, lapho abakhwabanisi bekhiqiza imibhalo engamanga kanye nobuso kusuka kudatha yangempela, okwenza ukukhwabanisa kuqiniseke futhi kube nzima ukukubona; kanye nokukhohliswa kwezithombe ozishuthe zona, lapho idokhumenti yangempela ihlanganiswa nesithombe esikhiqizwe i- deepfake ukuze kudlule amasistimu okubona ubuso. Lokhu kukhwabanisa kungenzeka ezindaweni ezihlukahlukene ohambweni lwedijithali, njengalapho ubhalisa amakhasimende amasha, ushintsha amadivayisi noma amaphasiwedi, futhi ucela imikhiqizo emisha nekhredithi, isibonelo.
Ukudala izixazululo ezisebenzayo zokuphepha kwedijithali kuyinkimbinkimbi njengokuvimbela ukukhwabanisa — ikakhulukazi uma sicabanga ukuthi imakethe yase-Brazilian inezinto ezithile, njengamamodeli ahlukahlukene omakhalekhukhwini nezinhlelo zokusebenza, amadivaysi eselula amadala asetshenziswayo, kanye nengxenye yabantu abanokufinyelela ku-inthanethi okulinganiselwe, okuvimbela ukuqaliswa kobuchwepheshe bezokuphepha obuthuthukisiwe.
Nokho, ngisho naphakathi kobunzima, kubalulekile ukuqinisekisa izinga eliphezulu lokuvikelwa kubakhwabanisi abahlale bethuthukisa amasu abo; ngakho-ke, izinkampani eziningi seziqalile ukuhlola amathuluzi azo zisebenzisa izindlela esezivele zisetshenziswe abakhwabanisi, njengamamaski e-2D kanye ne-3D, ukulingisa ubuso nokuzama ukweqa izinhlelo zokuqinisekisa. Ngaphezu kwalokho, ukudinga izitifiketi eziqinisekisa ukuthi ukuqinisekiswa kwe-biometric okusetshenzisiwe kusebenza kahle ekutholeni ama-deepfakes—njenge-iBeta 2 seal—kubalulekile ezinkampanini ukuthi zamukele ubuchwepheshe obuthembekile nobuvikelekile.
Kodwa-ke, ukuqinisekiswa kwe-biometric kukodwa akwanele ukuthola ama-deepfakes: indlela enezendlalelo eziningi iyadingeka. Ukuze uqinisekise ubuqiniso bedatha yomsebenzisi ngokunemba okukhulu, lobu buchwepheshe kufanele buhlanganiswe nezinye izinsiza, njengokuskena kwedokhumenti, i-OCR (ukubonwa kohlamvu olubonakalayo), nokuhlola kwangemuva . Ukuhlanganiswa kwalezi zinsiza zokuqinisekisa kungavimbela umsebenzisi ukuthi amukeleke enqubweni yokugibela esebenzisa idatha engamanga noma imibhalo yomunye umuntu, isibonelo.
Ngokuthuthukiswa kwamathuluzi e-AI akhiqizayo kanye nezindlela eziyinkimbinkimbi ezenza ukukhwabanisa kube lula futhi kushibhile ukwenza, ukukhwabanisa okuvela ku-deepfakes kuvame ukudlondlobala nakakhulu, kusuka ekubeni ngekho emthethweni kuya kumazinga "okuthengisa". Kulesi simo, izinkampani zidinga ukutshala imali ngokushesha ngangokunokwenzeka ezixazululweni ezixhuma ubuchwepheshe, ezishintshayo, kanye nobuhlakani, zikhethe izixazululo ezimaphakathi ezihlanganisa konke ukubhaliswa kwabasebenzisi, idokhumentari, kanye nedatha ye-biometric endaweni eyodwa.
