Deepfake

Tsar ɗin dijital ya ga haɓaka fasahar ban sha'awa amma mai tayar da hankali ta deepfake. Wannan hanyar manipulation video, wanda aka gina akan AI mai ƙirƙira, tana ƙirƙirar abun ciki na bidiyo da sauti mai inganci amma gaba ɗaya na ƙarya. Deepfakes suna haifar da babban ƙalubale ga al'ummarmu, suna tambayar bambanci tsakanin gaskiya da ƙarya.

Ka yi tunanin bidiyon shugaban ƙasar Ukraine Volodymyr Zelensky, yana kira ga sojojinsa su mika wuya. Wannan bidiyon ƙarya ya haifar da babban rikici kafin a cire shi. Haka kuma, an yi amfani da bidiyoyin da suka shafi Donald Trump a lokacin zaɓen shugaban ƙasa na Amurka don yin tasiri ga ra'ayin jama'a. Waɗannan misalan suna nuna ƙarfin tasirin deepfakes.

Illolin suna da yawa. A doka, Faransa tana hukunta yada hotunan da suka shafi mutum ba tare da izini ba (abin da aka rubuta a cikin sashe na 226-1 na Dokar Penal) da kuma yada bidiyoyin da aka canza ba tare da an bayyana su a matsayin haka ba (abin da aka rubuta a cikin sashe na 226-8). Waɗannan laifukan na iya haifar da tara har zuwa 45,000 euros da kuma hukuncin gidan yari na shekara guda.

Yawaitar deepfakes na barazana ga amincewa da abun ciki na dijital. Tare da kayan aikin kamar FakeApp, ƙirƙirar su yanzu yana samuwa ga kowa. Wannan yana ƙara haɗarin ba da labarin karya da kuma cutar da suna. Saboda haka yana da muhimmanci a fahimci wannan fasahar don kare kai da kyau.

Mahimman abubuwa da za a tuna

• Deepfakes suna amfani da AI mai ƙirƙira don ƙirƙirar bidiyoyin da suka yi kama da gaskiya amma ƙarya ne
• Za su iya yin tasiri ga ra'ayin jama'a da kuma cutar da suna na mutane
• Faransa tana hukunta yada bidiyoyin da aka canza ba tare da an bayyana su a matsayin haka ba
• Dawowar kayan aikin ƙirƙirar deepfakes yana ƙara haɗarin
• Bambancin tsakanin gaskiya da ƙarya yana zama mai wahala a cikin zamanin dijital

Menene deepfake: ma'anar da asali

Deepfakes suna amfani da basirar wucin gadi don haɗa abun ciki na bidiyo ko sauti cikin fayilolin da suka riga sun kasance. Wannan haɗin yana ƙirƙirar abun ciki mai yaudara ko na ƙarya. Kalmar “deepfake” tana haɗa “koyo mai zurfi” da “ƙarya”, tana nuna asalin sa a cikin AI.

Haɓakar GAN a 2014

A 2014, Ian Goodfellow ya ƙirƙiri hanyoyin haɗin gwiwar ƙirƙira (GAN). Wannan ƙirƙira ta buɗe hanya ga deepfakes masu hankali. Fasahar tana dogara ne akan algorithms guda biyu: ɗaya yana ƙirƙirar ƙarya, ɗayan kuma yana gano su. Wannan ci gaba ya nuna farawa na sintirin fuska na zamani.

Farkon a Reddit a 2017

Farkon deepfakes sun bayyana a ƙarshen 2017 a Reddit, ta hanyar wani mai amfani mara suna, “Deepfakes”. Waɗannan bidiyoyin, galibi suna da alaƙa da batsa kuma suna nuna shahararrun mutane, sun ja hankalin mutane. A watan Janairu 2018, FakeApp ya sa ƙirƙirar da raba deepfakes ya zama mai sauƙi.

Ci gaban fasahar cikin sauri

Ci gaban deepfakes ya kasance cikin sauri. Bisa ga Deeptrace, adadin bidiyoyin ya karu, yana tashi daga 8,000 a 2018 zuwa 15,000 a 2019. Wannan karuwar ta haifar da damuwa game da haɗarin tasiri da ba da labarin karya. Google ya amsa a 2019 ta hanyar fitar da bayanan bidiyo 3,000 don haɓaka kayan aikin gano su.

Yadda deepfake ke aiki

Deepfake, wannan ƙarya mai kama da gaskiya, tana fitowa daga AI mai ƙirƙira. Suna amfani da dabaru masu wahala. A tsakiyar wannan al'amari akwai hanyoyin haɗin gwiwar neurons, ko GAN, da Ian Goodfellow ya ƙirƙira a 2014.

Hanyoyin haɗin gwiwar neurons

GAN suna kunshe da algorithms guda biyu masu gwagwarmaya: mai ƙirƙira yana ƙirƙirar hotuna na ƙarya, yayin da mai rarrabawa ke ƙoƙarin gano su a matsayin ƙarya. Wannan gwagwarmaya tana inganta ingancin ƙaryar a kowanne mataki.

Tsarin koyo na atomatik

Koyon GAN yana faruwa ta hanyar gasa mai ci gaba. Mai ƙirƙira yana ƙoƙarin yaudara mai rarrabawa, wanda, a ƙarshe, yana inganta ƙarfin gano sa. Wannan maimaitawar inganta yana da mahimmanci don ingancin deepfake mai ban mamaki.

Sintirin fuska da murya

Don ƙirƙirar fuskoki, samfuran suna dogara ne akan wuraren fuska. Sintirin murya yana ba da damar canza murya. Waɗannan ci gaban suna sa bambanci tsakanin gaskiya da ƙarya ya zama mai wahala ba tare da kayan aikin musamman ba.

Gano waɗannan canje-canje yana buƙatar algorithms na AI masu iya gano alamomin da ba a iya ganewa. Gaban wannan barazanar, ilimin jama'a game da haɗarin dijital yana da mahimmanci. Yana taimakawa wajen haɓaka tunanin duba bayanai na yau da kullum.

Ƙarin nau'ikan deepfake da ke akwai

Deepfakes suna rarrabuwa cikin nau'uka da yawa, kowanne yana amfani da manipulation video da sintirin fuska ta hanyoyi daban-daban. Waɗannan ci gaban fasaha suna ba da damar ƙirƙirar deepfakes waɗanda za su iya yaudara ido na mutum da sauƙi.

Deepfakes na bidiyo sune mafi yawan. Suna haɗa da maye gurbin fuskoki a cikin bidiyoyin da suka riga sun kasance, suna haifar da labarai na ƙarya masu ƙarfi. A 2019, an kiyasta kimanin 15,000 na irin waɗannan bidiyoyin suna yawo.

Deepfakes na sauti, a gefe guda, suna da sirri. Suna kwaikwayon muryoyi na gaske, suna amfani da su musamman a cikin zamba na kuɗi. Wani shahararren misali shine na wani ma'aikaci da ya canza dala miliyan 25 bayan wani taron bidiyo na ƙarya.

Ƙirƙirar haruffa gaba ɗaya na ƙarya yana zama wani nau'i. Waɗannan deepfakes na iya haifar da fuskoki da muryoyi waɗanda ba su wanzu a cikin gaskiya ba, suna sa bambanci tsakanin gaskiya da ƙarya ya zama mai wahala fiye da kowane lokaci.

Gano waɗannan deepfakes har yanzu babban ƙalubale ne. Wani algorithm da jami'ar Buffalo ta haɓaka yana nazarin hasken haske akan iris tare da daidaito na 94%. Duk da haka, wannan yana yiwuwa ne kawai ga hotuna masu dindindin inda waɗannan hasken suke bayyana.

Babban haɗarin deepfake ga al'umma

Deepfakes suna zama barazana mai karuwa ga al'ummarmu. Suna haifar da tambayoyi na ɗabi'a dangane da basirar wucin gadi. Hakanan suna ba da ƙalubale masu yawa wajen yaki da waɗannan ƙaryar cikin inganci. Mu duba manyan haɗarin da suke wakilta.

Manipulation na ra'ayin jama'a

Deepfakes na iya shafar ra'ayin jama'a ta hanyar da ba ta dace ba. Wani shahararren misali shine bidiyon da aka canza na Emmanuel Macron yana tattara shara. Wannan yana nuna ƙarfin tasirin siyasa. Bugu da ƙari, kashi 33% na Faransawa suna da wahalar bambance abun ciki na gaske daga abun ciki da AI ta ƙirƙira, wanda ke ƙara wannan haɗarin.

Zambace-zambace da satar shaidar

Deepfakes suna sauƙaƙa zamba masu wahala. Wani kamfani daga Hong Kong ya rasa dala miliyan 25 saboda wata zamba da ta yi amfani da deepfakes. FBI na ba da rahoton karuwar zambace-zambace da suka shafi deepfakes, wanda yawanci suna nufin matasa.

Pornafiya ba tare da izini ba

96% na bidiyoyin deepfake da ke kan layi suna da alaƙa da batsa, suna nufin shahararrun mata kamar Emma Watson. Wannan al'amari yana haifar da tambayoyi masu tsanani game da girmamawa ga sirri da hakkin mutum.

Tasirin kan tsarin dimokuradiyya

Deepfakes na barazana ga tsaro na zaɓe. A 2023, an ƙirƙiri bidiyon deepfake na Joe Biden don hana masu zaɓe daga New Hampshire. Wannan yana nuna haɗarin ga tsarin dimokuradiyya. Faransa ta sanya hukunci masu tsanani, har zuwa shekaru 7 na gidan yari da tara 100,000 € don zamba da ta shafi deepfakes.

Gaban waɗannan ƙalubale, Tarayyar Turai ta bayyana deepfakes a cikin dokar AI ta. Wannan yana ba da tsarin doka don magance waɗannan matsalolin. Yaki da deepfakes yana buƙatar hanyar ɗabi'a ta AI da kuma ƙarin kulawa daga kowa.

Shahararrun misalan deepfake a duniya

Deepfakes, waɗannan canje-canje na bidiyo masu kama da gaskiya, sun yi tasiri a duniya tare da shahararrun misalai. A watan Yuli 2023, wani bidiyo na yaudara da ke nuna Emmanuel Macron yana sanar da murabus sa ya yadu sosai. Wannan yana nuna ƙarfin ba da labarin karya na wannan fasahar.

A Amurka, Steve Kramer an hukunta shi da tara dala miliyan 6 saboda ƙirƙirar saƙon sauti na ƙarya na Joe Biden. Wannan lamari yana nuna haɗarin amfani da deepfakes a fannin siyasa.

Duniya ta nishadi ba ta tsira ba. An ga deepfakes na batsa na Taylor Swift miliyoyin lokuta a kan kafofin sada zumunta. Wasu shahararru kamar Robert Downey Jr., Tom Hanks da Margot Robbie sun kasance maƙasudin waɗannan ƙaryar masu kama da gaskiya.

Zambace-zambace da ke amfani da deepfakes suna ƙaruwa. A Hong Kong, wani ma'aikaci an yaudare shi ta hanyar bidiyon da aka canza, wanda ya haifar da satar dala miliyan 26. A Asiya, wani hanyar soyayya ta ƙarya da ke amfani da profailin deepfake ta sace dala miliyan 46 daga maza guda.

Gano deepfake: alamomi da hanyoyi

Gano deepfake yana wakiltar babban kalubale, saboda saurin ci gaban AI mai ƙirƙira. Hanyoyin gano suna ci gaba don gano waɗannan ƙaryar dijital, suna zama masu kama da gaskiya fiye da kowane lokaci.

Alamomin gani da za a gano

Wasu alamomin gani na iya bayyana kasancewar deepfake. Misali, motsin ido ko fuskokin da ba su dace ba na iya bayyana ƙarya. Masana daga Jami'ar Hull sun ƙirƙiri wata hanya bisa ga nazarin hasken haske a cikin ido, tare da daidaito na kimanin 70%.

Rashin daidaito a cikin sauti da murya

Hanyar Phoneme-Visème, da masana daga Stanford da California suka haɓaka, tana gano rashin daidaito tsakanin motsin leɓɓa da sauti. Wannan fasahar tana ba da damar gano rashin daidaito masu laushi a cikin deepfakes na sauti.

Kayan aikin gano ta atomatik

Yawancin kayan aikin AI ana ƙirƙira su don inganta gano deepfakes:

• Reality Defender: yana gano deepfakes a cikin nau'o'in kafofin watsa labarai tare da hanyar multi-model
• Sentinel: yana amfani da algorithms na AI don nazarin canje-canje na dijital
• Intel FakeCatcher: yana gano bidiyoyin ƙarya tare da daidaito na 96% ta hanyar nazarin jini

Duk da ci gaban waɗannan kayan aikin, gano deepfakes har yanzu babban ƙalubale ne. Ci gaban ci gaba na deepfakes yana sa aikin ya zama mai wahala. Kulawa da amfani da hanyoyi da yawa suna da mahimmanci don gano waɗannan ƙaryar dijital da inganci.

Musamman deepfake na sauti

Deepfakes na sauti suna nuna ci gaba mai mahimmanci a fannin deepfakes masu hankali. Suna bambanta da sintirin fuska ta hanyar mai da hankali kan kwaikwayo na murya daga rikodi. Wannan fasahar tana da ƙwarewa wajen kama asalin murya na mutum, gami da intonations da saurin sa.

Tasirin deepfakes na sauti yana da yawa. Wani bidiyon deepfake da ke nuna Barack Obama tare da kalmomin da suka jawo cece-kuce ya ja hankalin mutane kusan miliyan 10. Wannan yana nuna ƙarfin waɗannan abun ciki wajen yaduwa cikin sauri.

Aikace-aikacen deepfakes na sauti sun wuce kwaikwayo. Suna buɗe hanya ga ƙirƙirar haruffa na wucin gadi kamar Lil Miquela, tare da muryoyi na wucin gadi masu ƙarfi. Wannan na iya juyar da hulɗa da nishadi na dijital.

Gaban wannan barazanar, ana haɓaka kayan aikin gano AI. Waɗannan fasahohin suna nufin gano abun ciki na sauti na wucin gadi. Suna yiwuwa su taimaka wajen yaki da ba da labarin karya na sauti. Duk da haka, gasa tsakanin ƙirƙira da gano tana ci gaba da kasancewa mai ƙarfi, tana bayyana muhimmancin kulawar jama'a.

Matakan kariya daga deepfake

Barazanar deepfakes na karuwa, yana haifar da sabbin hanyoyi don yaki da waɗannan canje-canje. Sabbin dabaru suna bayyana don kare mutane da ƙungiyoyi. Suna nufin dakatar da haɗarin wannan fasahar.

Hanyoyin fasaha

Sabbin fasahohi suna ƙirƙira don gano da dakatar da deepfakes:

• Kayan aikin tabbatar da inganci (tabbatar da matakai biyu, biometric na hali)
• Takardar shaidar abun ciki ta hanyar blockchain
• Tsarin gano ta AI wanda ke gano rashin daidaito masu laushi
• Kariya ga bayanai ta hanyar ɓoyewa mai zurfi

Tsarin doka da ƙa'idoji

Dokoki suna bayyana a cikin ƙasashe da yawa don tsara ƙirƙira da yada deepfakes. Waɗannan dokokin suna nufin sa masu ƙirƙira su zama masu alhakin da kuma kare waɗanda aka cuta.

Ilmantar da jama'a

Ilmantarwa yana da mahimmanci a cikin yaki da deepfakes. An ƙirƙiri shirye-shiryen horo don:

• Ilmantar da haɗarin deepfakes
• Koya yadda za a gano bidiyoyin da aka canza
• Haɓaka tunanin tunani game da abun ciki na kan layi

Ta hanyar haɗa waɗannan hanyoyin, al'umma na iya kare kanta daga haɗarin deepfakes. Wannan yana ba da damar kiyaye ci gaban fasahar dijital da ɗabi'a ta AI.

Rawar da kafofin sada zumunta ke takawa a fuskantar deepfake

Kafofin sada zumunta suna da mahimmanci a cikin yaki da deepfakes. Tare da karuwar lokutan zamba da suka shafi wannan fasahar, sun ƙara ƙoƙarinsu. Manufarsu ita ce kare masu amfani da su da kuma kiyaye ingancin bayanan da aka raba.

Ka'idojin tsarawa

Manyan kamfanonin yanar gizo sun karɓi tsare-tsare masu tsauri kan deepfakes. Wallafa abun ciki da aka canza na iya haifar da hukunci masu tsanani. Waɗannan hukuncin na iya kaiwa har zuwa shekaru 2 na gidan yari da tara 45,000 euros. Waɗannan matakan suna mai da hankali musamman kan dandamali kamar YouTube da TikTok, inda yada deepfakes ya fi yawa.

Tsarin gano

Gano deepfakes yana dogara ne akan fasahohi masu ci gaba. Dandamali suna nazarin motsin ido da ba su dace ba da kuma rashin ƙananan hawan ido. Hakanan suna gano fuskokin da ba su dace ba da rashin daidaito na jiki. Matsalolin daidaiton leɓɓa da hasken da ba su dace ba suna kuma bayyana a matsayin alamomi na canji.

Haɗin gwiwa tare da masu duba gaskiya

Don yaki da deepfakes, kafofin sada zumunta suna aiki tare da masu duba gaskiya. Wannan haɗin gwiwa yana ba da damar duba cikin sauri abun ciki mai shakku. Yana taimakawa wajen rage yaduwar bayanan ƙarya. Duk da haka, alhakin masu amfani yana da mahimmanci a cikin yaki da ba da labarin karya a kan kafofin sada zumunta.

Tasirin deepfake akan kamfanoni

Deepfake na zama barazana mai karuwa ga kamfanoni. Wannan fasahar ta manipulation video tana bayyana kamfanoni ga haɗarin kuɗi da suna masu yawa. A 2020, wata zamba da ta yi amfani da deepfake na murya ta ba da damar satar dala miliyan 35 daga United Arab Emirates, tana nuna girman haɗarin.

Har yanzu ana karuwar hare-haren yanar gizo ta deepfake, tare da ƙaruwa na 13% da VMware ta lura a 2022. Waɗannan hare-haren suna mai da hankali kan imel, saƙonnin wayar hannu da kafofin sada zumunta. Masu zamba suna amfani da bidiyoyi (58%) ko abun ciki na sauti (42%) don yaudara ma'aikata da samun damar tsarin bayanai masu mahimmanci.

Gaban wannan barazanar, kamfanoni suna ƙarfafa ɗabi'a ta AI da kuma kafa dabarun tsaro. Horar da ma'aikata, tabbatar da hanyoyin shiga biyu da kuma amfani da hanyoyin tsaro na yanar gizo masu ci gaba suna da mahimmanci. Haɗin gwiwa tare da ƙwararru masu shaidar, kamar waɗanda aka ba da suna na ExpertCyber, yana ba da damar kare kai daga waɗannan sabbin nau'ikan canje-canje na bidiyo.

Tambayoyi

Menene deepfake?

Deepfake shine ƙirƙira ta dijital mai kama da gaskiya, wanda ke fitowa daga basirar wucin gadi. Yana canza ko ƙirƙirar abun ciki ta hanyar amfani da hanyoyin haɗin gwiwar neurons (GAN). Waɗannan suna haɗa fuskoki, muryoyi, da motsi, suna sa bambanci tsakanin gaskiya da ƙarya ya zama mai wahala sosai.

Ta yaya fasahar deepfakes ke aiki?

Fasahar deepfakes tana dogara ne akan koyo mai zurfi da hanyoyin haɗin gwiwar neurons. Tana nazarin manyan bayanai don koya yadda za a ƙirƙiri abun ciki mai kama da gaskiya. Hanyar “mai ƙirƙira” tana ƙirƙirar abun ciki na ƙarya, yayin da hanyar “mai rarrabawa” ke ƙoƙarin gano ƙaryar. Wannan maimaitawar yana inganta ingancin deepfakes a kowanne mataki.

Menene manyan haɗarin deepfakes ga al'umma?

Deepfakes suna haifar da haɗari masu yawa, ciki har da manipulation na ra'ayin jama'a, zamba, da ƙirƙirar pornafiya ba tare da izini ba. Hakanan suna iya shafar zaɓe da rage amincewa da kafofin watsa labarai. Waɗannan haɗarin suna da mahimmanci ga dimokuradiyya da al'umma a gaba ɗaya.

Ta yaya za a gano deepfake?

Gano deepfakes yana buƙatar lura da alamomin gani da sauti. Motsin ido da ba su dace ba, fuskokin da ba su dace ba, rashin daidaito na sauti/hoto, da kuma rashin daidaito na murya suna daga cikin alamomin da ke da mahimmanci. Hakanan akwai kayan aikin gano ta atomatik, amma ingancinsu yana bambanta da ci gaban fasahar.

Wane matakai aka ɗauka don yaki da deepfakes?

Yaki da deepfakes yana haɗa da dabaru da dama. Haɓaka hanyoyin fasaha don tabbatar da ingancin abun ciki yana da mahimmanci. Kafa tsarin doka da ƙa'idoji ma yana da muhimmanci. Ilmantar da jama'a da aiwatar da tsare-tsare na tsarawa daga kafofin sada zumunta suna taka muhimmiyar rawa a cikin wannan yaki.

Shin deepfakes na sauti suna bambanta da deepfakes na bidiyo?

Eh, deepfakes na sauti suna da keɓancewa na musamman. Suna mai da hankali kan kwaikwayo na murya na gaske, gami da intonations da saurin sa. Waɗannan deepfakes suna da haɗari sosai saboda suna da wahalar ganowa fiye da deepfakes na bidiyo, musamman a cikin zambace-zambace na waya masu wahala.

Menene tasirin deepfakes akan kamfanoni?

Deepfakes suna haifar da haɗari masu yawa ga kamfanoni. Zasu iya cutar da suna, suyi tasiri ga kasuwannin kuɗi, da kuma haifar da leƙen asiri na masana'antu. Kamfanoni suna buƙatar haɓaka dabarun kariya da gudanar da rikice-rikice don fuskantar wannan barazana.

Ta yaya kafofin sada zumunta ke sarrafa barazanar deepfakes?

Kafofin sada zumunta suna ɗaukar dabaru daban-daban don yaki da deepfakes. Suna kafa tsare-tsare masu tsauri na tsarawa, suna haɓaka tsarin gano ta atomatik, da kuma haɗin gwiwa tare da masu duba gaskiya na waje. Manufarsu ita ce daidaita 'yancin faɗar albarkacin baki tare da buƙatar kare masu amfani daga ba da labarin karya.

Menene aikace-aikacen kyawawa na deepfakes?

Deepfakes suna da aikace-aikace masu kyau, ciki har da a cikin masana'antar nishadi don ƙirƙirar tasirin musamman. Hakanan ana amfani da su a cikin ilimi don yin kwaikwayo na tarihi da kuma a cikin fasaha don ƙirƙirarrun sabbin abubuwa. Fasahar na iya yin fassara fina-finai a cikin harsuna daban-daban cikin hanya mai kyau.

Ta yaya zan iya kare kaina daga deepfakes?

Don kare kanka, haɓaka tunanin ka na duba gaskiya da koyaushe duba tushen bayanan ka. Yi amfani da kayan aikin duba gaskiya idan zai yiwu kuma ka kasance a cikin shahararrun ci gaban deepfakes. Ilmantar da kai game da kafofin watsa labarai da ilimin dijital yana da mahimmanci don tunkarar wannan yanayin kafofin watsa labarai mai canzawa.