Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ekuphepheni Kwe-Cyber?

Isingeniso

I-AI Ekhiqizayo – izinhlelo zobuhlakani bokwenziwa ezikwazi ukudala okuqukethwe okusha noma izibikezelo – ziyavela njengamandla aguqulayo ekuphepheni kwe-cyber. Amathuluzi afana ne-OpenAI's GPT-4 abonise ikhono lokuhlaziya idatha eyinkimbinkimbi nokukhiqiza umbhalo ofana nowabantu, okwenza izindlela ezintsha zokuzivikela ezinsongweni ze-cyber. Ochwepheshe bezokuphepha kwe-cyber kanye nabenzi bezinqumo zebhizinisi kuzo zonke izimboni bahlola ukuthi i-AI ekhiqizayo ingaqinisa kanjani ukuzivikela ekuhlaselweni okuguqukayo. Kusukela kwezezimali kanye nokunakekelwa kwempilo kuya ekuthengisweni nasekuhulumeni, izinhlangano kuzo zonke izinhlaka zibhekene nemizamo yobugebengu bokweba imininingwane ebucayi, i-malware, kanye nezinye izinsongo i-AI ekhiqizayo engasiza ekuzilweni. Kuleli phepha elimhlophe, sihlola ukuthi i-AI ekhiqizayo ingasetshenziswa kanjani ekuphepheni kwe-cyber, sigqamisa izinhlelo zokusebenza zangempela, amathuba esikhathi esizayo, kanye nezinto ezibalulekile okufanele zicatshangelwe.

I-AI ekhiqizayo ihlukile ku-AI yokuhlaziya yendabuko ngokungagcini nje ngokuthola amaphethini kodwa nokudala okuqukethwe - kungaba ukulingisa ukuhlaselwa ukuze kuqeqeshwe ukuzivikela noma ukukhiqiza izincazelo zolimi lwemvelo zedatha yokuphepha eyinkimbinkimbi. Leli khono elikabili lenza libe yinkemba esika nhlangothi zombili: linikeza amathuluzi amasha anamandla okuzivikela, kodwa abenzi bezinsongo bangayisebenzisa futhi. Izigaba ezilandelayo zihlola uhla olubanzi lwamacala okusetshenziswa kwe-AI ekhiqizayo ekuphepheni kwe-cyber, kusukela ekuzivumelaniseni nokutholwa kwe-phishing kuya ekuthuthukiseni impendulo yezehlakalo. Siphinde sixoxe ngezinzuzo ezithenjiswa yilezi zindlela ezintsha ze-AI, kanye nezingozi (ezifana "nokungaqondi kahle" kwe-AI noma ukusetshenziswa kabi kokuphikisana) izinhlangano okumele zikuphathe. Okokugcina, sinikeza izindlela ezisebenzayo zokusiza amabhizinisi ukuthi ahlole futhi ahlanganise i-AI ekhiqizayo ngokuzibophezela emasu abo okuphepha kwe-cyber.

I-AI Ekhiqizayo ku-Cybersecurity: Ukubuka Konke

I-AI Ekhiqizayo ekuphepheni kwe-cyber ibhekisela kumamodeli e-AI - ngokuvamile amamodeli olimi amakhulu noma amanye amanethiwekhi e-neural - angakhiqiza ukuqonda, izincomo, ikhodi, noma ngisho nedatha yokwenziwa ukusiza emisebenzini yokuphepha. Ngokungafani namamodeli okubikezela kuphela, i-AI ekhiqizayo ingalingisa izimo futhi ikhiqize imiphumela efundeka kubantu (isb. imibiko, izexwayiso, noma ngisho namasampula ekhodi anonya) ngokusekelwe kudatha yayo yokuqeqesha. Leli khono lisetshenziswa ukubikezela , ukuthola, nokuphendula ezinsongweni ngezindlela ezinamandla kakhulu kunangaphambili (Iyini i-AI Ekhiqizayo Ekuphepheni Kwe-Cyber? - Amanethiwekhi e-Palo Alto). Isibonelo, amamodeli akhiqizayo angahlaziya izingodo ezinkulu noma izindawo zokugcina ubuhlakani bezinsongo futhi akhiqize isifinyezo esifushane noma isenzo esinconywayo, esebenza cishe njengo "msizi" we-AI emaqenjini okuphepha.

Ukuqaliswa kokuqala kwe-AI yokukhiqiza yokuvikela i-cyber kubonise ithemba. Ngo-2023, iMicrosoft yethule i-Security Copilot, umsizi osebenza nge-GPT-4 wabahlaziyi bezokuphepha, ukusiza ukuhlonza ukwephulwa kwemithetho nokuhlunga izinqubo ze-Microsoft zezimpawu ezingama-trillion angu-65 nsuku zonke (i-Microsoft Security Copilot ingumsizi omusha we-GPT-4 AI wokuphepha kwe-cyber | I-Verge). Abahlaziyi bangakhuthaza lolu hlelo ngolimi lwemvelo (isb. "Fingqa zonke izehlakalo zokuphepha emahoreni angu-24 edlule"), futhi umqhubi uzokhiqiza isifinyezo sendaba esiwusizo. Ngokufanayo, i-Treat Intelligence AI isebenzisa imodeli yokukhiqiza ebizwa ngokuthi i-Gemini ukuze ivumele usesho lwengxoxo ngesizindalwazi esikhulu se-intel se-threat Intel se-Google, ihlaziya ngokushesha ikhodi esolisayo futhi ifinyeze okutholakele ukusiza abazingeli be-malware (I-AI Yokukhiqiza Ingasetshenziswa Kanjani Ku-Cybersecurity? Izibonelo Eziyi-10 Zangempela). Lezi zibonelo zibonisa amandla: i-AI yokukhiqiza ingagaya idatha eyinkimbinkimbi, enkulu yokuphepha kwe-cyber futhi iveze ukuqonda ngendlela efinyelelekayo, isheshise ukwenza izinqumo.

Ngesikhathi esifanayo, i-AI ekhiqizayo ingadala okuqukethwe okungelona iqiniso kakhulu, okuyinzuzo yokulingisa nokuqeqesha (futhi, ngeshwa, kubahlaseli abenza ubunjiniyela bezenhlalo). Njengoba siqhubeka nezimo ezithile zokusetshenziswa, sizobona ukuthi ikhono le-AI ekhiqizayo lokuhlanganisa nokuhlaziya ulwazi lisekela izinhlelo zayo eziningi zokuphepha kwe-cyber. Ngezansi, singena ezimeni zokusetshenziswa ezibalulekile, sihlanganisa konke kusukela ekuvimbeleni i-phishing kuya ekuthuthukisweni kwesofthiwe okuvikelekile, ngezibonelo zendlela ngayinye esetshenziswa ngayo kuzo zonke izimboni.

Izicelo Eziyinhloko ze-AI Ekhiqizayo ku-Cybersecurity

Umfanekiso: Amacala abalulekile okusetshenziswa kwe-AI ekhiqizayo ekuphepheni kwe-cyber afaka phakathi ama-copilot e-AI amaqembu okuphepha, ukuhlaziywa kobungozi bekhodi, ukutholwa kwezinsongo eziguquguqukayo, ukulingisa ukuhlasela kwe-zero-day, ukuphepha kwe-biometric okuthuthukisiwe, kanye nokutholwa kwe-phishing (6 Amacala Okusetshenziswa kwe-AI Ekhiqizayo ekuphepheni kwe-Cyber ​​​​[+ Izibonelo] ).

Ukutholwa Nokuvimbela Ubugebengu Bokweba Imininingwane Ye-inthanethi

Ubugebengu bokweba imininingwane buhlala bungenye yezinsongo ze-cyber ezisakazeke kakhulu, bukhohlisa abasebenzisi ukuthi bachofoze izixhumanisi ezinonya noma baveze iziqinisekiso. I-AI Ekhiqizayo isetshenziswa kokubili ukuthola imizamo yobugebengu bokweba imininingwane futhi iqinise ukuqeqeshwa kwabasebenzisi ukuze kuvinjelwe ukuhlaselwa okuphumelelayo. Ohlangothini lokuzivikela, amamodeli e-AI angahlaziya okuqukethwe kwe-imeyili kanye nokuziphatha kwabathumeli ukuze abone izimpawu ezicashile zobugebengu bokweba imininingwane ezingase zingatholakali yizihlungi ezisekelwe emithethweni. Ngokufunda kumasethi amakhulu edatha ama-imeyili asemthethweni noma akhohlisayo, imodeli ekhiqizayo ingamaka ukungalingani ngethoni, amagama, noma umongo okhombisa ukukhwabanisa - ngisho nalapho uhlelo lolimi kanye nesipelingi kungasakuvezi. Eqinisweni, abacwaningi be-Palo Alto Networks baphawula ukuthi i-AI ekhiqizayo ingakhomba "izimpawu ezicashile zama-imeyili obugebengu bokweba imininingwane okungenzeka ukuthi azibonakali," kusiza izinhlangano ukuthi zihlale ziphambi kwabakhwabanisi (Iyini i-AI Ekhiqizayo Ekuphepheni Kwe-Cyber? - Palo Alto Networks).

Amaqembu ezokuphepha nawo asebenzisa i-AI yokukhiqiza ukuze alingise ukuhlaselwa kwe-phishing ukuze aqeqeshwe futhi ahlaziywe. Isibonelo, i-Ironscales yethule ithuluzi lokulingisa i-phishing elisebenzisa i-GPT elikhiqiza ngokuzenzakalelayo ama-imeyili e-phishing mbumbulu aklanyelwe abasebenzi benhlangano (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zezwe Langempela). Lawa ma-imeyili enziwe nge-AI abonisa amaqhinga akamuva abahlaseli, enikeza abasebenzi umkhuba ongokoqobo wokubona okuqukethwe kwe-phishy. Ukuqeqeshwa okunjalo okwenziwe ngezifiso kubalulekile njengoba abahlaseli ngokwabo bemukela i-AI ukuze bakhe izithiyo ezikholisayo. Okuphawulekayo, ngenkathi i-AI yokukhiqiza ingakhiqiza imiyalezo ye-phishing ephucukile kakhulu (seziphelile izinsuku zesiNgisi esiphukile esibonakala kalula), abavikeli bathole ukuthi i-AI ayinqobeki. Ngo-2024, abacwaningi be-IBM Security benze ucwaningo oluqhathanisa ama-imeyili e-phishing abhalwe ngabantu nalawo akhiqizwe yi-AI, futhi "okumangazayo ukuthi ama-imeyili akhiqizwe yi-AI ayesalula ukuwathola naphezu kohlelo lolimi olufanele" (6 Amacala Okusebenzisa i-AI Ekhiqizayo Ku-Cybersecurity [+ Izibonelo] ). Lokhu kusikisela ukuthi ukuqonda komuntu kuhlanganiswe nokutholwa okusizwa yi-AI kusengabona ukungalingani okucashile noma amasignali e-metadata kuma-scams abhalwe yi-AI.

I-AI Ekhiqizayo isiza ukuzivikela kwe-phishing ngezinye izindlela, futhi. Amamodeli angasetshenziswa ukukhiqiza izimpendulo ezenzakalelayo noma izihlungi ezihlola ama-imeyili asolisayo. Isibonelo, uhlelo lwe-AI lungaphendula i-imeyili ngemibuzo ethile ukuqinisekisa ubuqiniso bomthumeli noma lusebenzise i-LLM ukuhlaziya izixhumanisi nezinamathiselo ze-imeyili ebhokisini lesihlabathi, bese lufingqa noma iyiphi inhloso embi. Ipulatifomu yokuphepha ye-NVIDIA iMorpheus ikhombisa amandla e-AI kule nkundla - isebenzisa amamodeli e-NLP akhiqizayo ukuhlaziya ngokushesha nokuhlukanisa ama-imeyili, futhi kutholakale ukuthi ithuthukisa ukutholwa kwe-imeyili ye-spear-phishing ngo- 21% uma kuqhathaniswa namathuluzi okuphepha endabuko (6 Sebenzisa Amacala e-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). I-Morpheus iphinda iphrofayele amaphethini okuxhumana kwabasebenzisi ukuthola ukuziphatha okungavamile (njengomsebenzisi othumela i-imeyili ngokuzumayo amakheli amaningi angaphandle), okungabonisa i-akhawunti esengozini ethumela ama-imeyili e-phishing.

Empeleni, izinkampani kuzo zonke izimboni seziqala ukuthemba i-AI ukuthi ihlunge i-imeyili kanye nethrafikhi yewebhu ukuze kuhlaselwe ubunjiniyela bezenhlalo. Izinkampani zezimali, isibonelo, zisebenzisa i-AI ekhiqizayo ukuskena ukuxhumana ukuze kutholakale imizamo yokuzenza umuntu ongayena ongaholela ekukhohlisweni kwe-wire, kuyilapho abahlinzeki bezempilo besebenzisa i-AI ukuvikela idatha yesiguli ekuphulweni okuhlobene nokuphangwa kwe-phishing. Ngokukhiqiza izimo ezingokoqobo zokuphangwa kwe-phishing kanye nokuhlonza izimpawu zemiyalezo enonya, i-AI ekhiqizayo inezela ungqimba olunamandla kumasu okuvimbela ukuphangwa kwe-phishing. Okufanele ukwenze: I-AI ingasiza ekutholeni nasekuqedeni ukuhlaselwa kwe-phishing ngokushesha nangokunembile, njengoba abahlaseli besebenzisa ubuchwepheshe obufanayo ukuze bathuthukise umdlalo wabo.

Ukutholwa Kwe-Malware Nokuhlaziywa Kwezinsongo

I-malware yesimanje ishintsha njalo - abahlaseli bakhiqiza izinhlobo ezintsha noma ikhodi yokufiphaza ukuze badlule amasignesha e-antivirus. I-AI Ekhiqizayo inikeza amasu amasha kokubili ukuthola i-malware nokuqonda ukuziphatha kwayo. Enye indlela ukusebenzisa i-AI ukukhiqiza "amawele amabi" e-malware: abacwaningi bezokuphepha bangondla isampula ye-malware eyaziwayo kumodeli yokukhiqiza ukuze bakhe izinhlobo eziningi eziguquliwe zaleyo malware. Ngokwenza kanjalo, balindele ngempumelelo izinguquko umhlaseli angase azenze. Lezi zinhlobo ezikhiqizwa yi-AI zingasetshenziswa ukuqeqesha izinhlelo zokuthola i-antivirus kanye nokungena ngaphakathi, ukuze ngisho nezinguqulo eziguquliwe ze-malware zibonwe endle (6 Sebenzisa Amacala E-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Leli cebo elisebenzayo lisiza ukuphula umjikelezo lapho abaphangi beshintsha khona i-malware yabo ukuze bagweme ukutholwa futhi abavikeli kumele balwele ukubhala amasignesha amasha isikhathi ngasinye. Njengoba kuphawuliwe kwenye i-podcast yemboni, ochwepheshe bezokuphepha manje basebenzisa i-AI ekhiqizayo ukuze "balingise ithrafikhi yenethiwekhi futhi bakhiqize imithwalo enobungozi elingisa ukuhlaselwa okuyinkimbinkimbi," behlola ukuzivikela kwabo emndenini wonke wezinsongo kunokuba kube yisibonelo esisodwa. Lokhu kutholwa kosongo oluguquguqukayo kusho ukuthi amathuluzi okuphepha ayakwazi ukumelana ne-malware ehlukahlukene ebingadlula ngaphandle kwalokho.

Ngaphandle kokutholwa, i-AI yokukhiqiza isiza ekuhlaziyweni kwe-malware kanye nobunjiniyela obubuyela emuva, okuvame ukuba yimisebenzi enzima kakhulu kubahlaziyi bezinsongo. Amamodeli amakhulu ezilimi anganikwa umsebenzi wokuhlola ikhodi noma izikripthi ezisolisayo nokuchaza ngolimi olulula ukuthi ikhodi ihloselwe ukwenzani. Isibonelo sangempela yi -VirusTotal Code Insight, isici se-VirusTotal se-Google esisebenzisa imodeli ye-AI yokukhiqiza (i-Sec-PaLM ye-Google) ukukhiqiza izifinyezo zolimi lwemvelo zekhodi engaba yingozi (Ingasetshenziswa Kanjani i-AI Yokukhiqiza Ku-Cybersecurity? Izibonelo Eziyi-10 Zangempela Zomhlaba). Ngokuyisisekelo "uhlobo lwe-ChatGPT olunikezelwe ekubhaleni ikhodi yokuphepha," esebenza njengomhlaziyi we-malware we-AI osebenza amahora angama-24 ngosuku, izinsuku eziyi-7 ngesonto ukusiza abahlaziyi babantu baqonde izinsongo (6 Sebenzisa Amacala e-AI Yokukhiqiza Ku-Cybersecurity [+ Izibonelo] ). Esikhundleni sokufunda iskripthi esingajwayelekile noma ikhodi ye-binary, ilungu leqembu lokuphepha lingathola incazelo esheshayo evela ku-AI - isibonelo, "Lesi skripthi sizama ukulanda ifayela kusuka kuseva ye-XYZ bese sishintsha izilungiselelo zesistimu, okubonisa ukuziphatha kwe-malware." Lokhu kusheshisa kakhulu impendulo yezehlakalo, njengoba abahlaziyi bengakwazi ukuhlola nokuqonda i-malware entsha ngokushesha kunanini ngaphambili.

I-AI Ekhiqizayo isetshenziswa futhi ukukhomba i-malware kumasethi edatha amakhulu. Izinjini ze-antivirus zendabuko ziskena amafayela ukuthola amasignesha aziwayo, kodwa imodeli ekhiqizayo ingahlola izici zefayela futhi ibikezele ukuthi inonya yini ngokusekelwe kumaphethini afundiwe. Ngokuhlaziya izimfanelo zezigidigidi zamafayela (anonya futhi angenabungozi), i-AI ingase ibambe inhloso enonya lapho kungekho khona isignesha ecacile. Isibonelo, imodeli ekhiqizayo ingamaka i-executable njengesolisayo ngoba iphrofayili yayo yokuziphatha "ibukeka" njengohlobo oluthile lwe-ransomware eyalubona ngesikhathi sokuqeqeshwa, yize i-binary iyintsha. Lokhu kutholwa okusekelwe ekuziphatheni kusiza ukulwa ne-malware entsha noma yezinsuku ezingenalutho. I-Threat Intelligence AI ye-Google (ingxenye ye-Chronicle/Mandiant) kubikwa ukuthi isebenzisa imodeli yayo ekhiqizayo ukuhlaziya ikhodi engaba nonya futhi "isize ochwepheshe bezokuphepha ngempumelelo nangendlela ephumelelayo ekulweni ne-malware nezinye izinhlobo zezinsongo." (I-AI Ekhiqizayo Ingasetshenziswa Kanjani Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela).

Ngakolunye uhlangothi, kumele sivume ukuthi abahlaseli bangasebenzisa i-AI yokukhiqiza lapha futhi - ukudala ngokuzenzakalelayo i-malware ezivumelanisa nezimo. Eqinisweni, ochwepheshe bezokuphepha baxwayisa ngokuthi i-AI yokukhiqiza ingasiza izigebengu ze-cyber ukuthi zakhe i-malware okunzima ukuyibona (Iyini i-AI Ekhiqizayo ku-Cybersecurity? - Amanethiwekhi e-Palo Alto). Imodeli ye-AI ingayalwa ukuthi iguqule ingxenye ye-malware ngokuphindaphindiwe (ishintshe isakhiwo sayo sefayela, izindlela zokubethela, njll.) ize igweme konke ukuhlolwa kwe-antivirus okwaziwayo. Lokhu kusetshenziswa kokuphikisana kuyinkinga ekhulayo (ngezinye izikhathi ebizwa ngokuthi “i-malware enikwe amandla yi-AI” noma i-malware ye-polymorphic njengesevisi). Sizoxoxa ngezingozi ezinjalo kamuva, kodwa kugcizelela ukuthi i-AI yokukhiqiza iyithuluzi kulo mdlalo wekati negundane osetshenziswa ngabavikeli nabahlaseli.

Sekukonke, i-AI yokukhiqiza ithuthukisa ukuzivikela kwe-malware ngokuvumela amaqembu okuphepha ukuthi acabange njengomhlaseli - edala izinsongo ezintsha nezixazululo ngaphakathi. Kungakhathaliseki ukuthi ikhiqiza i-malware yokwenziwa ukuze ithuthukise amazinga okutholwa noma isebenzisa i-AI ukuchaza nokuqukatha i-malware yangempela etholakala kumanethiwekhi, lawa masu asebenza kuzo zonke izimboni. Ibhange lingasebenzisa ukuhlaziywa kwe-malware okuqhutshwa yi-AI ukuhlaziya ngokushesha i-macro esolisayo kuspredishithi, kuyilapho inkampani yokukhiqiza ingase ithembele ku-AI ukuthola i-malware eqondisa izinhlelo zokulawula zezimboni. Ngokwandisa ukuhlaziywa kwe-malware yendabuko nge-AI yokukhiqiza, izinhlangano zingaphendula emikhankasweni ye-malware ngokushesha nangokushesha kunangaphambili.

Ukuhlaziywa Kobunhloli Nokuzilawula Kobungozi

Nsuku zonke, izinhlangano zihlaselwa ngedatha yobuhlakani bezinsongo - kusukela kokuphakelayo kwezinkomba ezisanda kutholakala zokuyekethisa (ama-IOC) kuya emibikweni yabahlaziyi mayelana namaqhinga amasha abaduni. Inselelo yamaqembu okuphepha ukuhlunga lolu lwazi oluningi nokukhipha ukuqonda okusebenzayo. I-AI ekhiqizayo ibalulekile ekuzenzekeleni ukuhlaziywa nokusetshenziswa kobuhlakani bezinsongo. Esikhundleni sokufunda ngesandla imibiko eminingi noma okufakiwe kwedatha, abahlaziyi bangasebenzisa i-AI ukufingqa nokufaka umongo we-intelligence yezinsongo ngesivinini somshini.

Isibonelo esisodwa esiqondile yi-suite ye-Treat Intelligence ka-Google , ehlanganisa i-AI ekhiqizayo (imodeli ye-Gemini) nenqwaba yedatha ye-Google yosongo evela ku-Mandiant kanye ne-VirusTotal. Le AI inikeza "ukusesha okuxoxisanayo kuyo yonke indawo enkulu ye-Google yolwazi lwezosongo" , okuvumela abasebenzisi ukuthi babuze imibuzo yemvelo mayelana nezinsongo futhi bathole izimpendulo ezihluziwe ( Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela ). Isibonelo, umhlaziyi angabuza, "Sike sabona noma iyiphi i-malware ehlobene ne-Treatment Group X eqondise embonini yethu?" futhi i-AI izodonsa ulwazi olufanele, mhlawumbe iphawule ukuthi "Yebo, i-Treatment Group X ixhunywe nomkhankaso wobugebengu bokweba imininingwane ebucayi ngenyanga edlule isebenzisa i-malware Y" , kanye nesifinyezo sokuziphatha kwaleyo malware. Lokhu kunciphisa kakhulu isikhathi sokuqoqa ulwazi obekungadinga ukubuza amathuluzi amaningi noma ukufunda imibiko emide.

I-AI Ekhiqizayo ingaphinde ihlanganise futhi ifingqe izitayela zokusongela. Ingase ihlole izinkulungwane zezikhala zebhulogi yezokuphepha, izindaba zokuphulwa komthetho, kanye nezingxoxo zewebhu ezimnyama bese ikhiqiza isifinyezo esiphezulu "sezinsongo eziphezulu ze-cyber kuleli sonto" ukuze kufinyelelwe i-CISO. Ngokwesiko, leli zinga lokuhlaziya nokubika lidinga umzamo omkhulu womuntu; manje imodeli ehlelwe kahle ingayiloba ngemizuzwana, abantu belungisa kuphela umphumela. Izinkampani ezifana ne-ZeroFox zithuthukise i-FoxGPT, ithuluzi le-AI elikhiqizayo elenzelwe ngqo "ukusheshisa ukuhlaziywa nokufingqa kobuhlakani kuwo wonke amasethi amakhulu edatha," kufaka phakathi okuqukethwe okunonya kanye nedatha yokuphanga (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zangempela). Ngokuzenzakalela ukuphakanyiswa okunzima kokufunda kanye nedatha yokubhekisela, i-AI ivumela amaqembu e-intel okusongelayo ukuthi agxile ekwenzeni izinqumo kanye nokuphendula.

Esinye isimo sokusetshenziswa ukuzingela izinsongo ezingxoxweni. Cabanga ngomhlaziyi wezokuphepha exhumana nomsizi we-AI: “Ngibonise noma yiziphi izimpawu zokukhishwa kwedatha emahoreni angu-48 edlule” noma “Yiziphi izisulu ezintsha ezisengozini ezisetshenziswa kakhulu kuleli sonto?” I-AI ingahumusha umbuzo, iseshe amalogi angaphakathi noma imithombo yangaphandle ye-Intel, futhi iphendule ngempendulo ecacile noma ngisho nohlu lwezehlakalo ezifanele. Lokhu akulona iphutha - izinhlelo zesimanje zolwazi lokuphepha kanye nokuphathwa kwemicimbi (SIEM) seziqala ukufaka imibuzo yolimi lwemvelo. I-suite yezokuphepha ye-QRadar ye-IBM, isibonelo, ingeza izici ze-AI ezikhiqizayo ngo-2024 ukuze ivumele abahlaziyi “babuze […] imibuzo ethile mayelana nendlela yokuhlasela efingqiwe” yesigameko futhi bathole izimpendulo ezinemininingwane. Ingaphinde “ihumushe futhi ifingqe ubuhlakani bezinsongo obufanele kakhulu” ngokuzenzakalelayo (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zangempela Zomhlaba). Empeleni, i-AI ekhiqizayo iguqula izintaba zedatha yobuchwepheshe zibe ukuqonda okulinganiselwe kwengxoxo ngokufunwa.

Kuyo yonke imboni, lokhu kunemiphumela emikhulu. Umhlinzeki wezokunakekelwa kwempilo angasebenzisa i-AI ukuze ahlale enolwazi ngamaqembu akamuva e-ransomware aqondise ezibhedlela, ngaphandle kokunikeza umhlaziyi ucwaningo lwesikhathi esigcwele. I-SOC yenkampani yokuthengisa ingafingqa ngokushesha amaqhinga amasha e-POS malware lapho itshela abasebenzi be-IT esitolo. Futhi kuhulumeni, lapho idatha yosongo evela ezinhlanganweni ezahlukahlukene kufanele ihlanganiswe khona, i-AI ingakhiqiza imibiko ehlanganisiwe eqokomisa izixwayiso ezibalulekile. Ngokuzenzakalelayo ukuqoqwa nokuchazwa kolwazi losongo, i-AI ekhiqizayo isiza izinhlangano ukuthi zisabele ngokushesha ezinsongweni ezivelayo futhi inciphise ingozi yokulahlekelwa izixwayiso ezibalulekile ezifihliwe emsindweni.

Ukuthuthukiswa Kwesikhungo Sokusebenza Kwezokuphepha (i-SOC)

Izikhungo Zokusebenza Kwezokuphepha zidume ngokukhathala kokuxwayisa kanye nenani elikhulu ledatha. Umhlaziyi we-SOC ojwayelekile angase adlule ezinkulungwaneni zezexwayiso nemicimbi nsuku zonke, ephenya izehlakalo ezingaba khona. I-AI Ekhiqizayo isebenza njengesiphindaphindi samandla kuma-SOC ngokwenza umsebenzi ojwayelekile ngokuzenzakalelayo, ukunikeza izifinyezo ezihlakaniphile, ngisho nokuhlela ezinye izimpendulo. Umgomo ukwenza ngcono ukuhamba komsebenzi kwe-SOC ukuze abahlaziyi babantu bakwazi ukugxila ezindabeni ezibaluleke kakhulu kuyilapho umshayeli we-AI ephatha okunye.

Uhlelo lokusebenza olukhulu ukusebenzisa i-AI ekhiqizayo njenge- "Analyst's Copilot". I-Microsoft Security Copilot, okuphawulwe ngayo ngaphambili, iyisibonelo salokhu: "iklanyelwe ukusiza umsebenzi womhlaziyi wezokuphepha kunokuwushintsha," isiza ekuphenyweni kwezehlakalo kanye nokubika (i-Microsoft Security Copilot ingumsizi omusha we-GPT-4 AI wezokuphepha kwe-cyber | The Verge). Empeleni, lokhu kusho ukuthi umhlaziyi angafaka idatha eluhlaza - amalogi e-firewall, umugqa wesikhathi womcimbi, noma incazelo yesigameko - bese ecela i-AI ukuthi iyihlaziye noma iyifingqe. I-copilot ingase ikhiphe indaba efana nokuthi, "Kubonakala sengathi ngo-2:35 AM, ukungena okusolisayo okuvela ku-IP X kuphumelele ku-Server Y, kulandelwe ukudluliswa kwedatha okungavamile, okubonisa ukwephulwa okungenzeka kwaleyo seva." Lolu hlobo lokuhlanganiswa komongo ngokushesha lubaluleke kakhulu lapho isikhathi sibaluleke kakhulu.

Abashayeli bezindiza be-AI nabo bayasiza ekunciphiseni umthwalo wokuhlola izinga loku-1. Ngokusho kwedatha yemboni, ithimba lezokuphepha lingachitha amahora ayi-15 ngesonto lihlela izexwayiso ezingaba ngu-22,000 kanye nemiphumela engamanga (6 Sebenzisa Amacala Okukhiqiza I-AI ku-Cybersecurity [+ Izibonelo] ). Nge-AI ekhiqizayo, eziningi zalezi zexwayiso zingahlolwa ngokuzenzakalelayo - i-AI ingalahla lezo ezisobala ukuthi azinangozi (ngokucabanga okunikeziwe) futhi igqamise lezo ezidinga ukunakwa ngempela, ngezinye izikhathi iphakamise nokuthi yini eza kuqala. Eqinisweni, amandla e-AI ekhiqizayo ekuqondeni umongo asho ukuthi ingahlanganisa izexwayiso ezingase zibonakale zingenangozi zodwa kodwa ndawonye zibonise ukuhlaselwa kwezigaba eziningi. Lokhu kunciphisa ithuba lokuphuthelwa ukuhlaselwa ngenxa "yokukhathala okuqaphile."

Abahlaziyi be-SOC basebenzisa nolimi lwemvelo ne-AI ukusheshisa ukuzingela kanye nophenyo. ye-Purple AI , isibonelo, ihlanganisa isikhombimsebenzisi esisekelwe ku-LLM nedatha yokuphepha yesikhathi sangempela, okuvumela abahlaziyi ukuthi "babuze imibuzo eyinkimbinkimbi yokuzingela izinsongo ngesiNgisi esilula futhi bathole izimpendulo ezisheshayo nezinembile" (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zezwe Langempela). Umhlaziyi angathayipha athi, "Ingabe kukhona ama-endpoints axhumaniswe ne-domain badguy123[.]com ngenyanga edlule?", futhi i-Purple AI izosesha ngamalogi ukuze iphendule. Lokhu kusindisa umhlaziyi ekubhaleni imibuzo noma izikripthi zedatha - i-AI ikwenza ngaphansi kwe-hood. Kusho nokuthi abahlaziyi abancane bangaphatha imisebenzi eyayidinga unjiniyela onolwazi ngezilimi zemibuzo ngaphambilini, ethuthukisa ngempumelelo ithimba ngosizo lwe-AI. Ngempela, abahlaziyi babika ukuthi isiqondiso se-AI yokukhiqiza "sithuthukisa amakhono abo kanye nekhono labo", njengoba abasebenzi abancane manje sebengathola ukwesekwa kokubhala amakhodi noma amathiphu okuhlaziya avela ku-AI lapho kudingeka, okunciphisa ukuthembela ekuceleni njalo amalungu eqembu eliphezulu usizo (6 Sebenzisa Amacala e-AI Yokukhiqiza ku-Cybersecurity [+ Izibonelo] ).

Okunye ukulungiswa kwe-SOC ukufingqa kwezehlakalo okuzenzakalelayo kanye nokubhala imibhalo. Ngemva kokuthi isigameko sesisingathiwe, othile kumele abhale umbiko - umsebenzi abaningi abawuthola uyisicefe. I-AI ekhiqizayo ingathatha idatha ye-forensic (amalogi esistimu, ukuhlaziywa kwe-malware, umugqa wesikhathi wezenzo) futhi ikhiqize umbiko wezehlakalo wokuqala. I-IBM yakha leli khono ku-QRadar ukuze "ngokuchofoza okukodwa" isifinyezo sesehlakalo sikhiqizwe kubabambiqhaza abahlukene (abaphathi, amaqembu e-IT, njll.) (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela). Lokhu akugcini nje ngokusindisa isikhathi kodwa futhi kuqinisekisa ukuthi akukho lutho olunganakwa embikweni, njengoba i-AI ingafaka yonke imininingwane efanele njalo. Ngokufanayo, ukuze kulandelwe imithetho kanye nokuhlolwa, i-AI ingagcwalisa amafomu noma amathebula obufakazi ngokusekelwe kudatha yezehlakalo.

Imiphumela yangempela iyathandeka. Abasebenzisa i-SOAR yokuqala eqhutshwa yi-Swimlane eqhutshwa yi-AI (ukuhlelwa kwezokuphepha, ukuzenzekela, kanye nokuphendula) babika izinzuzo ezinkulu zokukhiqiza - i-Global Data Systems, isibonelo, ibone ithimba labo le-SecOps liphatha umthwalo omkhulu kakhulu wamacala; omunye umqondisi uthe "engikwenzayo namuhla ngabahlaziyi abangu-7 cishe kuzothatha amalungu abasebenzi abangu-20 ngaphandle" kwe-automation esebenzisa i-AI (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber). Ngamanye amazwi, i-AI ku-SOC ingandisa umthamo. Kuyo yonke imboni, kungakhathaliseki ukuthi yinkampani yezobuchwepheshe ebhekene nezaziso zokuphepha kwamafu noma izinhlelo zokuqapha isitshalo sokukhiqiza ze-OT, amaqembu e-SOC azokwazi ukuthola ukutholwa okusheshayo kanye nempendulo, izigameko ezimbalwa ezingabanjwanga, kanye nemisebenzi ephumelelayo ngokwamukela abasizi be-AI abakhiqizayo. Kumayelana nokusebenza ngokuhlakanipha - ukuvumela imishini ukuthi iphathe imisebenzi ephindaphindwayo nesindayo yedatha ukuze abantu bakwazi ukusebenzisa umuzwa wabo kanye nobuchwepheshe lapho kubaluleke kakhulu.

Ukuphathwa Kokusengozini kanye Nokulingisa Usongo

Ukuhlonza nokuphatha ubuthakathaka – ubuthakathaka kusofthiwe noma izinhlelo abahlaseli abangazisebenzisa – kuwumsebenzi oyinhloko wokuphepha kwe-inthanethi. I-AI ekhiqizayo ithuthukisa ukuphathwa kobuthakathaka ngokusheshisa ukutholakala, isize ekubekeni phambili ama-patch, ngisho nokulingisa ukuhlaselwa kwalezo zibuthakathaka ukuze kuthuthukiswe ukulungela. Empeleni, i-AI isiza izinhlangano ukuthola nokulungisa izimbobo ezisezivikeloni zazo ngokushesha, futhi ihlola ukuzivikela ngaphambi kokuba abahlaseli bangempela benze kanjalo.

Esinye isicelo esibalulekile ukusebenzisa i-AI yokukhiqiza ukuze kubuyekezwe ikhodi ezenzakalelayo kanye nokutholwa kobungozi. Ama-codebase amakhulu (ikakhulukazi izinhlelo ze-legend) avame ukuba namaphutha okuphepha angabonwa. Amamodeli e-AI yokukhiqiza angaqeqeshwa emikhubeni evikelekile yokufaka amakhodi kanye namaphethini avamile eziphazamisi, bese ekhishwa kukhodi yomthombo noma kuma-binary ahlanganisiwe ukuthola ubungozi obungaba khona. Isibonelo, abacwaningi be-NVIDIA bathuthukise ipayipi le-AI yokukhiqiza elingahlaziya izitsha zesofthiwe ye-legend futhi libone ubuthakathaka "ngokunemba okuphezulu - kufika ku-4× ngokushesha kunochwepheshe babantu." (6 Sebenzisa Amacala e-AI Yokukhiqiza ku-Cybersecurity [+ Izibonelo] ). I-AI ngokuyisisekelo yafunda ukuthi ikhodi engavikelekile ibukeka kanjani futhi yakwazi ukuskena ngesofthiwe yamashumi eminyaka ukuze iveze imisebenzi eyingozi kanye nemitapo yolwazi, isheshisa kakhulu inqubo evamile ehamba kancane yokuhlola ikhodi ngesandla. Lolu hlobo lwethuluzi lungashintsha umdlalo wezimboni ezifana nezezimali noma uhulumeni othembele kuma-codebase amakhulu, amadala - i-AI isiza ukwenza ukuphepha kube kwesimanje ngokumba izinkinga abasebenzi abangathatha izinyanga noma iminyaka ukuzithola (uma kwenzeka).

I-AI Ekhiqizayo iphinde isize ekusebenzeni kokuphathwa kobungozi ngokucubungula imiphumela yokuskena kobungozi kanye nokuyibeka phambili. Amathuluzi afana ne -ExposureAI asebenzisa i-AI ekhiqizayo ukuvumela abahlaziyi ukuthi babuze idatha yobungozi ngolimi olulula futhi bathole izimpendulo ezisheshayo (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zezwe Langempela). I-ExposureAI ingakwazi "ukufingqa indlela ephelele yokuhlasela endabeni" yobungozi obuthile obubucayi, ichaza ukuthi umhlaseli angayihlanganisa kanjani nobunye ubuthakathaka ukuze abeke engcupheni uhlelo. Ize income nezenzo zokulungisa futhi iphendule imibuzo elandelayo mayelana nengozi. Lokhu kusho ukuthi lapho kumenyezelwa i-CVE entsha ebalulekile (Common Vulnerability and Exposures), umhlaziyi angabuza i-AI, "Ingabe kukhona amaseva ethu athintekile yile CVE futhi yisiphi isimo esibi kakhulu uma singayilungisi?" futhi athole ukuhlolwa okucacile okuthathwe kudatha yokuskena yenhlangano. Ngokufaka ubungozi kumongo (isb. lokhu kuvezwe ku-inthanethi nakuseva enenani eliphezulu, ngakho-ke kuyinto eza kuqala), i-AI ekhiqizayo isiza amaqembu ukulungisa ngokuhlakanipha ngezinsiza ezilinganiselwe.

Ngaphezu kokuthola nokuphatha ubuthakathaka obaziwayo, i-AI ekhiqizayo inegalelo ekuhlolweni kokungena kanye nokulingisa ukuhlasela - empeleni ukuthola obungaziwa noma ukuhlola izilawuli zokuphepha. Amanethiwekhi akhiqizayo aphikisanayo (ama-GAN), uhlobo lwe-AI ekhiqizayo, asetshenziswe ukudala idatha yokwenziwa elingisa ithrafikhi yenethiwekhi yangempela noma ukuziphatha komsebenzisi, okungafaka phakathi amaphethini okuhlasela afihliwe. Ucwaningo lwango-2023 luphakamise ukusebenzisa ama-GAN ukukhiqiza ithrafikhi yokuhlasela yangempela yosuku olulodwa ukuqeqesha izinhlelo zokuthola ukungena (6 Amacala Okusebenzisa i-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Ngokupha ama-IDS ngezimo zokuhlasela ezenziwe nge-AI (ezingafaki engcupheni yokusebenzisa i-malware yangempela kumanethiwekhi okukhiqiza), izinhlangano zingaqeqesha ukuzivikela kwazo ukuze zibone izinsongo ezintsha ngaphandle kokulinda ukuthi zishaywe ngokoqobo. Ngokufanayo, i-AI ingalingisa umhlaseli ohlola uhlelo - isibonelo, ezama ngokuzenzakalelayo amasu ahlukahlukene okusebenzisa endaweni ephephile ukuze abone ukuthi kukhona yini ophumelelayo. I-US Defense Advanced Research Projects Agency (DARPA) ibona isithembiso lapha: i-2023 AI Cyber ​​Challenge yayo isebenzisa ngokusobala i-AI ekhiqizayo (njengezinhlobo ezinkulu zezilimi) "ukuthola nokulungisa ngokuzenzakalelayo ubuthakathaka kusofthiwe yomthombo ovulekile" njengengxenye yomncintiswano ( I-DARPA Ihlose Ukuthuthukisa i-AI, Izicelo Zokuzimela Ama-Warfighters Angazethemba > UMnyango Wezokuvikela wase-US > Izindaba ZoMnyango Wezokuvikela ). Lesi sinyathelo sigcizelela ukuthi i-AI ayisizi nje kuphela ekulungiseni izikhala ezaziwayo; ithola ngenkuthalo ezintsha futhi iphakamisa izilungiso, umsebenzi ovame ukukhawulelwa kubacwaningi bezokuphepha abanekhono (nababizayo).

I-AI Ekhiqizayo ingadala ngisho nama-honeypot ahlakaniphile kanye namawele edijithali okuzivikela. Ama-startup athuthukisa izinhlelo zobuqili eziqhutshwa yi-AI ezilingisa ngendlela ekholisayo amaseva noma amadivayisi angempela. Njengoba omunye u-CEO echazile, i-AI ekhiqizayo ingakwazi "ukuhlanganisa izinhlelo zedijithali ukuze zilingise ezangempela futhi ihehe abaduni" (6 Sebenzisa Amacala E-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Lawa ma-honeypot akhiqizwa yi-AI aziphatha njengendawo yangempela (ake sithi, idivayisi ye-IoT mbumbulu ethumela i-telemetry evamile) kodwa ikhona kuphela ukuze ihehe abahlaseli. Lapho umhlaseli ehlose i-decoy, i-AI empeleni ibakhohlisile ukuthi bembule izindlela zabo, abavikeli abangazifunda futhi bazisebenzise ukuqinisa izinhlelo zangempela. Lo mqondo, oqhutshwa yimodeli yobunjiniyela, unikeza indlela ebheke phambili yokuguqula amatafula kubahlaseli, usebenzisa inkohliso ethuthukiswe yi-AI.

Kuzo zonke izimboni, ukuphathwa kobuthakathaka okusheshayo nokuhlakaniphile kusho ukwephulwa okuncane. Ku-IT yezempilo, isibonelo, i-AI ingase ibone ngokushesha umtapo wolwazi osephelelwe yisikhathi kudivayisi yezokwelapha bese ikhuthaza ukulungiswa kwe-firmware ngaphambi kokuba noma yimuphi umhlaseli ayisebenzise kabi. Ebhange, i-AI ingalingisa ukuhlaselwa kwangaphakathi kohlelo lokusebenza olusha ukuqinisekisa ukuthi idatha yamakhasimende ihlala iphephile ngaphansi kwazo zonke izimo. Ngakho-ke i-AI ekhiqizayo isebenza njenge-microscope kanye nomhloli wokucindezeleka wesimo sokuphepha kwezinhlangano: ikhanyisa amaphutha afihliwe kanye nezinhlelo zokucindezela ngezindlela zobuciko zokuqinisekisa ukuqina.

Ukukhiqizwa Kwekhodi Evikelekile Nokuthuthukiswa Kwesofthiwe

Amathalente e-AI akhiqizayo awagcini nje ekuboneni ukuhlaselwa - abuye andise ekudaleni izinhlelo eziphephile kakhulu kusukela ekuqaleni. Ekuthuthukisweni kwesofthiwe, abakhiqizi bekhodi ye-AI (njenge-GitHub Copilot, i-OpenAI Codex, njll.) bangasiza abathuthukisi ukuthi babhale ikhodi ngokushesha ngokuphakamisa iziqeshana zekhodi noma ngisho nemisebenzi yonke. I-engeli yokuphepha kwe-cyber iwukuqinisekisa ukuthi lezi zingcezu zekhodi eziphakanyisiwe ze-AI ziphephile futhi zisebenzisa i-AI ukuthuthukisa imikhuba yokubhala ikhodi.

Ngakolunye uhlangothi, i-AI ekhiqizayo ingasebenza njengomsizi wokubhala ikhodi ofaka imikhuba emihle yokuphepha. Abathuthukisi bangakhuthaza ithuluzi le-AI, "Khiqiza umsebenzi wokusetha kabusha iphasiwedi ku-Python," futhi kungcono ukuthola ikhodi engasebenzi nje kuphela kodwa futhi elandela iziqondiso eziphephile (isb. ukuqinisekiswa kokufaka okufanele, ukuloga, ukuphathwa kwamaphutha ngaphandle kolwazi oluvuzayo, njll.). Umsizi onjalo, oqeqeshwe ngezibonelo eziningi zekhodi evikelekile, angasiza ekunciphiseni amaphutha abantu aholela ekubuthakathakeni. Isibonelo, uma unjiniyela ekhohlwa ukuhlanza okokufaka komsebenzisi (ukuvula umnyango wokujova kwe-SQL noma izinkinga ezifanayo), i-AI ingakufaka lokho ngokuzenzakalelayo noma ibaxwayise. Amanye amathuluzi okubhala ikhodi ye-AI manje alungiswa ngedatha egxile kwezokuphepha ukuze afeze le njongo ngqo - empeleni, ukubhanqa izinhlelo ze-AI ngonembeza wezokuphepha.

Kodwa-ke, kukhona uhlangothi oluhlukile: i-AI ekhiqizayo ingaletha kalula ubuthakathaka uma ingalawulwa kahle. Njengoba uchwepheshe wezokuphepha weSophos uBen Verschaeren aphawulile, ukusebenzisa i-AI ekhiqizayo yokufaka ikhodi "kulungile ngekhodi emfushane, eqinisekisiwe, kodwa kuyingozi lapho ikhodi engahlolwanga ihlanganiswa" ezinhlelweni zokukhiqiza. Ingozi iwukuthi i-AI ingase ikhiqize ikhodi enembile ngokunengqondo engaphephile ngezindlela umuntu ongeyena uchwepheshe angase angaziboni. Ngaphezu kwalokho, abenzi bobubi bangathonya ngamabomu amamodeli e-AI omphakathi ngokuwatshala ngamaphethini ekhodi asengozini (uhlobo lokusha kwedatha) ukuze i-AI iphakamise ikhodi engaphephile. Iningi labathuthukisi ababona ochwepheshe bezokuphepha, ngakho-ke uma i-AI isikisela ikhambi elilula, bangase bayisebenzise ngokungacabangi, bengaqapheli ukuthi inephutha (6 Sebenzisa Amacala E-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Lokhu kukhathazeka kungokoqobo - empeleni, kukhona uhlu lwe-OWASP Top 10 manje lwama-LLM (amamodeli ezilimi ezinkulu) oluchaza izingozi ezivamile ezinjengalezi ekusebenziseni i-AI yokufaka ikhodi.

Ukuze balwisane nalezi zinkinga, ochwepheshe basikisela “ukulwa ne-AI yokukhiqiza nge-AI yokukhiqiza” emkhakheni wokubhala amakhodi. Empeleni, lokho kusho ukusebenzisa i-AI ukubuyekeza nokuhlola ikhodi ebhalwe abanye abantu be-AI (noma abantu). I-AI ingaskena ikhodi entsha ishesha kakhulu kunombuyekezi wekhodi yomuntu futhi iveze ubuthakathaka obungaba khona noma izinkinga ze-logic. Sesivele sibona amathuluzi avelayo ahlanganiswa nomjikelezo wokuphila kokuthuthukiswa kwesofthiwe: ikhodi ibhaliwe (mhlawumbe ngosizo lwe-AI), bese imodeli yokukhiqiza eqeqeshwe ezimisweni zekhodi evikelekile iyibuyekeza futhi ikhiqize umbiko wanoma yiziphi izinkinga (ake sithi, ukusetshenziswa kwemisebenzi elahliwe, ukuhlolwa kokuqinisekisa okungekho, njll.). Ucwaningo lwe-NVIDIA, okukhulunywe ngalo ngaphambili, olufinyelele ukutholwa kobuthakathaka okusheshayo okungu-4x kukhodi luyisibonelo sokusebenzisa i-AI ekuhlaziyweni kwekhodi ephephile (6 Amacala Okusebenzisa i-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ).

Ngaphezu kwalokho, i-AI ekhiqizayo ingasiza ekudaleni ukucushwa okuphephile kanye nezikripthi. Isibonelo, uma inkampani idinga ukufaka ingqalasizinda yefu evikelekile, unjiniyela angacela i-AI ukuthi ikhiqize izikripthi zokucushwa (Ingqalasizinda njengeKhodi) ngezilawuli zokuphepha (njengokuhlukaniswa kwenethiwekhi okufanele, izindima ze-IAM ezinelungelo eliphansi) ezifakwe kuzo. I-AI, njengoba iqeqeshwe ngezinkulungwane zokucushwa okunjalo, ingakhiqiza isisekelo unjiniyela alungisa ngaso. Lokhu kusheshisa ukusethwa okuphephile kwezinhlelo futhi kunciphisa amaphutha okulungiselelwa kabi - umthombo ovamile wezigameko zokuphepha kwamafu.

Ezinye izinhlangano zisebenzisa i-AI ekhiqizayo ukuze zilondoloze isisekelo solwazi lwamaphethini okubhala ikhodi aphephile. Uma unjiniyela engaqiniseki ukuthi angayisebenzisa kanjani isici esithile ngokuphephile, angabuza i-AI yangaphakathi efunde kumaphrojekthi enkampani adlule kanye neziqondiso zokuphepha. I-AI ingase ibuyisele indlela enconywayo noma ngisho nesiqephu sekhodi esivumelana nezidingo zokusebenza kanye nezindinganiso zokuphepha zenkampani. Le ndlela isetshenziswe amathuluzi afana ne- Secureframe's Questionnaire Automation, edonsa izimpendulo ezinqubomgomo zenkampani kanye nezixazululo zesikhathi esidlule ukuqinisekisa izimpendulo ezihambisanayo nezinembile (empeleni ekhiqiza amadokhumenti aphephile) (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zangempela Zomhlaba). Lo mqondo uhumusha ekubhaleni ikhodi: i-AI "ekhumbula" ukuthi usebenzise kanjani into ngokuphephile ngaphambili futhi ikuqondise ukuthi uyenze ngaleyo ndlela futhi.

Ngamafuphi, i-AI yokukhiqiza ithonya ukuthuthukiswa kwesofthiwe ngokwenza usizo lokufaka amakhodi oluvikelekile lutholakale kalula. Izimboni ezakha isofthiwe eningi eyenziwe ngokwezifiso - ubuchwepheshe, ezezimali, ezokuvikela, njll. - zingazuza ngokuba nabashayeli be-AI abangochwepheshe abasheshisa ukufaka amakhodi kuphela kodwa futhi abasebenza njengombuyekezi wezokuphepha oqaphile njalo. Uma kulawulwa kahle, la mathuluzi e-AI anganciphisa ukwethulwa kobuthakathaka obusha futhi asize amaqembu okuthuthukisa anamathele emikhubeni emihle kakhulu, noma ngabe iqembu lingenaye uchwepheshe wezokuphepha ohilelekile kuzo zonke izinyathelo. Umphumela uba isofthiwe eqinile kakhulu ekulweni nokuhlaselwa kusukela ngosuku lokuqala.

Ukusekelwa Kwempendulo Yezehlakalo

Uma kwenzeka isigameko sokuphepha kwe-inthanethi – kungaba ukuqubuka kwe-malware, ukwephulwa kwedatha, noma ukuphazamiseka kwesistimu ngenxa yokuhlaselwa – isikhathi sibalulekile. I-AI ekhiqizayo isetshenziswa kakhulu ukusekela amaqembu okuphendula izigameko (IR) ekuvimbeleni nasekulungiseni izigameko ngokushesha nangolwazi olwengeziwe. Umqondo uwukuthi i-AI ingathwala omunye wemithwalo yophenyo kanye neyokubhala ngesikhathi sesigameko, futhi iphakamise noma yenze izenzo ezithile zokuphendula zibe ngokuzenzakalela.

Indima ebalulekile ye-AI ku-IR ukuhlaziywa kwezehlakalo zesikhathi sangempela kanye nokufingqa. Phakathi kwesehlakalo, abaphenduli bangase badinge izimpendulo zemibuzo efana nokuthi “Umhlaseli ungene kanjani?”, “Yiziphi izinhlelo ezithintekile?”, kanye nokuthi “Yimuphi idatha engase ibe sengozini?”. I-AI ekhiqizayo ingahlaziya amalogi, izexwayiso, kanye nedatha ye-forensic evela ezinhlelweni ezithintekile futhi inikeze ngokushesha ukuqonda. Isibonelo, i-Microsoft Security Copilot ivumela umphenduli wezehlakalo ukuthi anikeze ubufakazi obuhlukahlukene (amafayela, ama-URL, amalogi emicimbi) futhi acele umugqa wesikhathi noma isifinyezo (i-Microsoft Security Copilot ingumsizi omusha we-GPT-4 AI wezokuphepha kwe-cyber | The Verge). I-AI ingaphendula ngokuthi: “Ukwephulwa kungenzeka kwaqala nge-imeyili yobugebengu bokweba imininingwane ebucayi kumsebenzisi uJohnDoe ngo-10:53 GMT equkethe i-malware X. Uma isiqediwe, i-malware yadala umnyango wangemuva owasetshenziswa ezinsukwini ezimbili kamuva ukuze uthuthele eceleni kuseva yezezimali, lapho iqoqa khona idatha.” Ukuba nalesi sithombe esihambisanayo ngemizuzu hhayi amahora kwenza ithimba likwazi ukwenza izinqumo ezinolwazi (njengokuthi yiziphi izinhlelo okufanele zihlukaniswe) ngokushesha okukhulu.

I-AI Ekhiqizayo ingaphakamisa nezenzo zokuvimbela nokulungisa. Isibonelo, uma indawo yokugcina ithelelekile yi-ransomware, ithuluzi le-AI lingakhiqiza iskripthi noma isethi yemiyalelo yokuhlukanisa lowo mshini, ukukhubaza ama-akhawunti athile, nokuvimba ama-IP anobungozi aziwayo ku-firewall - empeleni ukwenziwa kwe-playbook. I-Palo Alto Networks iphawula ukuthi i-AI ekhiqizayo iyakwazi "ukukhiqiza izenzo noma izikripthi ezifanele ngokusekelwe kuhlobo lwesigameko", ngokuzenzakalela izinyathelo zokuqala zokuphendula (Iyini i-AI Ekhiqizayo ku-Cybersecurity? - I-Palo Alto Networks). Esimweni lapho ithimba lezokuphepha likhungathekile (ake sithi ukuhlasela okusabalele kumadivayisi amaningi), i-AI ingase yenze ngqo ezinye zalezi zenzo ngaphansi kwezimo ezivunyelwe kusengaphambili, isebenze njengomphenduli omncane osebenza ngokungakhathali. Isibonelo, i-ejenti ye-AI ingasetha kabusha ngokuzenzakalelayo iziqinisekiso ezibona ukuthi zisengozini noma zihlukanise ababungazi ababonisa umsebenzi ononya ofana nephrofayili yesigameko.

Ngesikhathi sokuphendula ezigamekweni, ukuxhumana kubalulekile - kokubili ngaphakathi kwethimba kanye nakubabambiqhaza. I-AI ekhiqizayo ingasiza ngokubhala imibiko yokubuyekezwa kwezigameko noma izifinyezo ngokushesha. Esikhundleni sokuba unjiniyela ayeke ukuxazulula izinkinga zakhe ukuze abhale isibuyekezo se-imeyili, angabuza i-AI, “Fingqa okwenzekile kulesi sigameko kuze kube manje ukuze azise abaphathi.” I-AI, ngemva kokungenisa idatha yesigameko, ingaveza isifinyezo esifushane: “Kusukela ngo-3 ntambama, abahlaseli sebefinyelele ama-akhawunti abasebenzisi amabili kanye namaseva ama-5. Idatha ethintekile ifaka amarekhodi amaklayenti kudathabheyisi X. Izinyathelo zokuvimba: Ukufinyelela i-VPN yama-akhawunti asengozini kuhoxisiwe futhi amaseva ahlukanisiwe. Izinyathelo ezilandelayo: ukuskena noma yiziphi izindlela zokuphikelela.” Umphenduli angaqinisekisa noma alungise lokhu ngokushesha futhi akuthumele, eqinisekisa ukuthi ababambe iqhaza bagcinwa benolwazi olunembile, olusesikhathini.

Ngemva kokuba uthuli seluthulile, ngokuvamile kuba nombiko wesigameko onemininingwane okufanele uwulungiselele kanye nezifundo ezifundwe okufanele zihlanganiswe. Lena enye indawo lapho ukwesekwa kwe-AI kukhanya khona. Ingabuyekeza yonke idatha yesigameko futhi ikhiqize umbiko wangemva kwesigameko ohlanganisa imbangela, ukulandelana kwezikhathi, umthelela, kanye nezincomo. I-IBM, isibonelo, ihlanganisa i-AI ekhiqizayo ukuze idale "izifinyezo ezilula zamacala okuphepha kanye nezehlakalo ezingabelwana ngazo nabathintekayo" ngokucindezela inkinobho (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela). Ngokwenza kube lula ukubika ngemuva kwesenzo, izinhlangano zingenza ngokushesha ukuthuthukiswa futhi zibe nemibhalo engcono ngezinjongo zokuthobela imithetho.

Okunye ukusetshenziswa okusha okubheke phambili ukulingisa izigameko okuqhutshwa yi-AI. Njengoba umuntu angenza kanjani i-fire drill, ezinye izinkampani zisebenzisa i-AI ekhiqizayo ukuze zidlule ezimweni zezehlakalo "ezizokwenzeka uma". I-AI ingalingisa indlela i-ransomware engasakazeka ngayo uma kubhekwa ukwakheka kwenethiwekhi, noma ukuthi umuntu wangaphakathi angakhipha kanjani idatha, bese ethola amaphuzu okusebenza kahle kwezinhlelo zamanje zokuphendula. Lokhu kusiza amaqembu ukulungiselela nokulungisa izincwadi zokudlala ngaphambi kokuba kwenzeke isigameko sangempela. Kufana nokuba nomeluleki wempendulo yezehlakalo ohlala ethuthuka ohlola ukulungela kwakho njalo.

Ezimbonini ezisengozini enkulu njengezezimali noma ezempilo, lapho isikhathi sokungasebenzi noma ukulahleka kwedatha ngenxa yezigameko kubiza kakhulu, la makhono e-IR aqhutshwa yi-AI ayakhanga kakhulu. Isibhedlela esibhekene nesigameko se-inthanethi asikwazi ukukhokhela ukungasebenzi kwesikhathi eside kwesistimu - i-AI esiza ngokushesha ekulawuleni ingase isindise impilo. Ngokufanayo, isikhungo sezezimali singasebenzisa i-AI ukusingatha ukuhlolwa kokuqala kokungenelela okusolwayo kokukhwabanisa ngo-3 ekuseni, ukuze kuthi ngesikhathi abantu abaku-inthanethi sebeku-inthanethi, kube sekuqediwe umsebenzi omningi (ukuvala ama-akhawunti athintekile, ukuvimba ukuthengiselana, njll.). Ngokwandisa amaqembu okuphendula ezigamekweni nge-AI yokukhiqiza, izinhlangano zinganciphisa kakhulu izikhathi zokuphendula futhi zithuthukise ukunemba kokuphathwa kwazo, ekugcineni zinciphise umonakalo ovela ezigamekweni ze-inthanethi.

Ukuhlaziywa Kokuziphatha Nokutholwa Kwe-Anomaly

Ukuhlaselwa okuningi kwe-cyber kungabanjwa ngokuqaphela lapho okuthile kuphambuka ekuziphatheni "okuvamile" - kungaba yi-akhawunti yomsebenzisi elanda inani elingavamile ledatha noma idivayisi yenethiwekhi exhumana ngokuzumayo nomphathi ongamazi. I-Generative AI inikeza amasu athuthukile okuhlaziya ukuziphatha kanye nokutholwa kwe-anomaly, ukufunda amaphethini ajwayelekile abasebenzisi nezinhlelo bese kuthi umaka lapho okuthile kubukeka kabi.

Ukutholwa kwe-anomaly yendabuko kuvame ukusebenzisa imingcele yezibalo noma ukufunda okulula komshini kuma-metric athile (ukukhuphuka kokusetshenziswa kwe-CPU, ukungena ngemvume ngamahora angajwayelekile, njll.). I-AI Ekhiqizayo ingaqhubeka nalokhu ngokudala amaphrofayili okuziphatha anezici eziningi. Isibonelo, imodeli ye-AI ingadla ukungena ngemvume, amaphethini okufinyelela amafayela, kanye nemikhuba ye-imeyili yesisebenzi ngokuhamba kwesikhathi futhi yakhe ukuqonda okunezici eziningi "okujwayelekile" kwalowo msebenzisi. Uma leyo akhawunti kamuva yenza okuthile okungaphandle kwenqubo yayo (njengokungena ngemvume usuka ezweni elisha nokufinyelela amafayela amaningi e-HR phakathi kwamabili), i-AI izobona ukuphambuka hhayi nje ku-metric eyodwa kodwa njengephethini yokuziphatha ephelele engahambisani nephrofayili yomsebenzisi. Ngokwezobuchwepheshe, amamodeli akhiqizayo (njengama-autoencoders noma amamodeli okulandelana) angabonisa ukuthi "okujwayelekile" kubukeka kanjani bese ekhiqiza ububanzi bokuziphatha obulindelekile. Lapho iqiniso liwela ngaphandle kwalelo banga, limakwe njenge-anomaly (Kuyini i-AI Ekhiqizayo Ku-Cybersecurity? - Palo Alto Networks).

Ukuqaliswa okusebenzayo kusekuqapheni ithrafikhi yenethiwekhi. Ngokusho kocwaningo lwango-2024, ama-54% ezinhlangano zase-US akhombe ukuqapha ithrafikhi yenethiwekhi njengesibonelo esisetshenziswa kakhulu se-AI ekuphepheni kwe-cyber (eNyakatho Melika: amacala okusetshenziswa kwe-AI aphezulu ekuphepheni kwe-cyber emhlabeni wonke ngo-2024). I-AI ekhiqizayo ingafunda amaphethini okuxhumana ajwayelekile enethiwekhi yebhizinisi - ukuthi yimaphi amaseva avame ukukhuluma wodwa, ukuthi yimaphi amanani edatha ahambayo phakathi namahora ebhizinisi uma kuqhathaniswa nobusuku bonke, njll. Uma umhlaseli eqala ukukhipha idatha kuseva, ngisho kancane ukuze agweme ukutholakala, uhlelo olusekelwe ku-AI lungaqaphela ukuthi "Iseva A ayithumeli nje kuphela i-500MB yedatha ngo-2 AM ku-IP yangaphandle" futhi iphakamise isexwayiso. Ngenxa yokuthi i-AI ayisebenzisi nje imithetho engaguquki kodwa iyimodeli eguqukayo yokuziphatha kwenethiwekhi, ingabamba ukungalingani okucashile imithetho engaguquki (njenge-"alert uma idatha > X MB") engase iphuthe noma iphawule ngephutha. Lolu hlobo oluguquguqukayo yilona olwenza ukutholwa kwe-anomaly okuqhutshwa yi-AI kube namandla ezindaweni ezifana namanethiwekhi okuthengiselana kwamabhange, ingqalasizinda yamafu, noma imikhumbi yedivayisi ye-IoT, lapho ukuchaza imithetho eqondile yokujwayelekile noma okungavamile kuyinkimbinkimbi kakhulu.

I-AI Ekhiqizayo iyasiza futhi ngokuhlaziywa kokuziphatha komsebenzisi (i-UBA), okuyisihluthulelo sokubona izinsongo zangaphakathi noma ama-akhawunti asengozini. Ngokudala isisekelo somsebenzisi noma inhlangano ngayinye, i-AI ingathola izinto ezifana nokusetshenziswa kabi kweziqinisekiso. Isibonelo, uma uBob kusuka ekubalweni kwezimali eqala ngokuzumayo ukubuza ngesizindalwazi samakhasimende (into angakaze ayenze ngaphambili), imodeli ye-AI yokuziphatha kukaBob izophawula lokhu njengokungavamile. Kungase kungabi yi-malware - kungaba yindaba yokuthi iziqinisekiso zikaBob zebiwe futhi zisetshenziswe umhlaseli, noma uBob aphenye lapho kungafanele khona. Noma ngabe yikuphi, ithimba lezokuphepha lithola ulwazi lokuphenya. Izinhlelo ze-UBA ezinjalo eziqhutshwa yi-AI zikhona emikhiqizweni ehlukahlukene yokuphepha, futhi amasu okukhiqiza amamodeli aphakamisa ukunemba kwawo futhi anciphise ama-alamu angamanga ngokucabangela umongo (mhlawumbe uBob usephrojekthini ekhethekile, njll., i-AI engayiphetha ngezinye izikhathi kusuka kwenye idatha).

Emkhakheni wobunikazi kanye nokuphathwa kokufinyelela, ukutholwa kwe-deepfake kuyisidingo esikhulayo - i-AI ekhiqizayo ingadala amazwi namavidiyo okwenziwa akhohlisa ukuphepha kwe-biometric. Ngokuthakazelisayo, i-AI ekhiqizayo ingasiza futhi ekutholeni lawa ma-deepfake ngokuhlaziya izinto ezicashile kumsindo noma kuvidiyo okunzima ngabantu ukuqaphela. Sibone isibonelo nge-Accenture, eyasebenzisa i-AI ekhiqizayo ukulingisa ukubonakaliswa kobuso okungenakubalwa kanye nezimo ukuqeqesha izinhlelo zayo ze-biometric ukuhlukanisa abasebenzisi bangempela kuma-deepfake akhiqizwe yi-AI. Eminyakeni emihlanu edlule, le ndlela yasiza i-Accenture ukuthi isuse amaphasiwedi angu-90% wezinhlelo zayo (ukuthuthela kuma-biometrics nezinye izici) futhi inciphise ukuhlaselwa ngo-60% (6 Sebenzisa Amacala e-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Empeleni, basebenzise i-AI ekhiqizayo ukuqinisa ukuqinisekiswa kwe-biometric, okwenza kube namandla ekuhlaselweni okukhiqizayo (umfanekiso omuhle wokulwa ne-AI ye-AI). Lolu hlobo lokulingisa ukuziphatha - kulokhu ukuqaphela umehluko phakathi kobuso bomuntu obuphilayo nobuso obuhlanganisiwe be-AI - lubalulekile njengoba sithembele kakhulu ku-AI ekuqinisekisweni.

Ukutholwa kwe-Anomaly okunikezwa amandla yi-AI yokukhiqiza kusebenza kuzo zonke izimboni: kwezempilo, ukuqapha ukuziphatha kwedivayisi yezokwelapha ukuze kutholakale izimpawu zokugenca; kwezezimali, ukuqapha izinhlelo zokuhweba ukuze kutholakale amaphethini angajwayelekile angabonisa ukukhwabanisa noma ukuphathwa kabi kwe-algorithmic; kwezamandla/izinsiza, ukubuka izimpawu zesistimu yokulawula ukuze kutholakale izimpawu zokungena. Ukuhlanganiswa kobubanzi (ukubheka zonke izici zokuziphatha) kanye nokujula (ukuqonda amaphethini ayinkimbinkimbi) okunikezwa yi-AI yokukhiqiza kwenza kube ithuluzi elinamandla lokubona izinkomba ze-needle-in-a-haystack zesigameko se-cyber. Njengoba izinsongo ziba zifihlakele, zicashe phakathi kwemisebenzi evamile, leli khono lokuchaza ngokunembile "okuvamile" nokumemeza lapho okuthile kuphazamiseka kuba yinto ebalulekile. Ngakho-ke i-AI yokukhiqiza isebenza njengomlindi ongakhathali, efunda njalo futhi ebuyekeza incazelo yayo yokujwayelekile ukuze ihambisane nezinguquko emvelweni, futhi ixwayisa amaqembu okuphepha ngeziphambeko ezifanele ukuhlolwa okuseduze.

Amathuba kanye nezinzuzo ze-AI ekhiqizayo ekuphepheni kwe-cyber

Ukusetshenziswa kwe-AI yokukhiqiza ekuphepheni kwe-cyber kuletha amathuba amaningi nezinzuzo ezinhlanganweni ezizimisele ukwamukela la mathuluzi. Ngezansi, sifingqa izinzuzo ezibalulekile ezenza i-AI yokukhiqiza ibe yinto ekhangayo ezinhlelweni zokuphepha kwe-cyber:

• Ukutholwa Nokuphendula Okusheshayo Kwengozi: Izinhlelo ze-AI ezikhiqizayo zingahlaziya idatha eningi ngesikhathi sangempela futhi zibone izinsongo ngokushesha kakhulu kunokuhlaziywa komuntu ngesandla. Le nzuzo yesivinini isho ukutholakala kokuhlaselwa kusenesikhathi kanye nokulawulwa kwezehlakalo okusheshayo. Empeleni, ukuqapha ukuphepha okuqhutshwa yi-AI kungabamba izinsongo ezingathatha abantu isikhathi eside ukuziqondanisa. Ngokuphendula ezigamekweni ngokushesha (noma ngisho nokwenza izimpendulo zokuqala ngokuzimela), izinhlangano zinganciphisa kakhulu isikhathi sokuhlala kwabahlaseli kumanethiwekhi azo, zinciphise umonakalo.

• Ukunemba Okuthuthukisiwe kanye Nokumbozwa Kwezinsongo: Ngenxa yokuthi bafunda njalo kudatha entsha, amamodeli akhiqizayo angazivumelanisa nezinsongo ezishintshayo futhi abambe izimpawu ezicashile zomsebenzi ononya. Lokhu kuholela ekuqondeni okuthuthukisiwe kokutholwa (okuncane okungalungile okungamanga kanye nokuhle okungamanga) uma kuqhathaniswa nemithetho engaguquki. Isibonelo, i-AI efunde izimpawu ze-imeyili yobugebengu bokweba imininingwane ebucayi noma ukuziphatha kwe-malware ingakhomba izinhlobo ezingakaze zibonwe ngaphambili. Umphumela uwukumbozwa okubanzi kwezinhlobo zezinsongo - kufaka phakathi ukuhlaselwa okusha - ukuqinisa ukuma kokuphepha okuphelele. Amaqembu okuphepha nawo athola ukuqonda okuningiliziwe ekuhlaziyweni kwe-AI (isb. izincazelo zokuziphatha kwe-malware), okuvumela ukuzivikela okunembe kakhulu nokuqondiwe (Kuyini i-AI Ekhiqizayo Ku-Cybersecurity? - Amanethiwekhi e-Palo Alto).

• Ukuzenzakalela Kwemisebenzi Ephindaphindwayo: I-AI Ekhiqizayo ihamba phambili ekwenzeni imisebenzi yokuphepha ejwayelekile, edinga abasebenzi abaningi - kusukela ekuhloleni izingodo nokuhlanganisa imibiko kuya ekubhaleni izikripthi zempendulo yezehlakalo. Lokhu kwenzakalela kunciphisa umthwalo kubahlaziyi babantu, kubakhulule ekugxileni isu eliphezulu kanye nokwenza izinqumo eziyinkimbinkimbi (Kuyini i-AI Ekhiqizayo ku-Cybersecurity? - Palo Alto Networks). Imisebenzi ebalulekile kodwa ebalulekile njengokuskena ubungozi, ukuhlolwa kokucushwa, ukuhlaziywa kwemisebenzi yomsebenzisi, kanye nokubika kokuhambisana kungaphathwa (noma okungenani kuqala) yi-AI. Ngokusingatha le misebenzi ngesivinini somshini, i-AI ayithuthukisi nje kuphela ukusebenza kahle kodwa futhi inciphisa amaphutha abantu (isici esibalulekile ekuphulweni kwemithetho).

• Ukuzivikela Okusebenzayo Nokulingisa: I-AI Ekhiqizayo ivumela izinhlangano ukuthi zishintshe kusukela ekuphepheni okusebenzayo ziye ekuphepheni okusebenzayo. Ngamasu afana nokulingisa ukuhlasela, ukukhiqizwa kwedatha yokwenziwa, kanye nokuqeqeshwa okusekelwe ezimweni, abavikeli bangalindela futhi balungiselele izinsongo ngaphambi kokuba zenzeke ezweni langempela. Amaqembu ezokuphepha angalingisa ukuhlaselwa kwe-cyber (imikhankaso yobugebengu bokweba imininingwane ebucayi, ukuqubuka kwe-malware, i-DDoS, njll.) ezindaweni eziphephile ukuze ahlole izimpendulo zawo futhi avikele noma yibuphi ubuthakathaka. Lokhu kuqeqeshwa okuqhubekayo, okuvame ukungenzeki ngokuphelele ngomzamo womuntu kuphela, kugcina ukuzivikela kubukhali futhi kusesikhathini. Kufana "nokujima komlilo" kwe-cyber - i-AI ingaphonsa izinsongo eziningi ezicatshangelwayo ekuzivikeleni kwakho ukuze ukwazi ukuzijwayeza futhi uthuthuke.

• Ukwandisa Ubuchwepheshe Babantu (i-AI njenge-Power Multiplier): I-AI Ekhiqizayo isebenza njengomhlaziyi osemncane ongakhathali, umeluleki, kanye nomsizi ohlanganiswe ndawonye. Inganikeza amalungu eqembu angenalo ulwazi oluningi isiqondiso nezincomo ezivame ukulindeleka kochwepheshe abanolwazi, ithuthukise ubuchwepheshe eqenjini lonke (6 Amacala Okusebenzisa i-AI Ekhiqizayo ku-Cybersecurity [+ Izibonelo] ). Lokhu kubaluleke kakhulu uma kubhekwa ukuntuleka kwamakhono ku-cybersecurity - i-AI isiza amaqembu amancane ukuthi enze okuningi ngokuncane. Ngakolunye uhlangothi, abahlaziyi abanolwazi bayazuza ekuphatheni i-AI ngokubhonga kanye nokuveza ukuqonda okungabonakali, abangakuqinisekisa bese bekwenza. Umphumela ophelele ithimba lezokuphepha elikhiqiza kakhulu futhi elinekhono, ne-AI ikhulisa umthelela welungu ngalinye lomuntu (Ingasetshenziswa Kanjani I-AI Ekhiqizayo ku-Cybersecurity).

• Ukusekelwa Nokubika Okuthuthukisiwe Kwezinqumo: Ngokuhumushela idatha yobuchwepheshe ekuqondeni kolimi lwemvelo, i-AI ekhiqizayo ithuthukisa ukuxhumana nokwenza izinqumo. Abaholi bezokuphepha bathola ukubonakala okucacile ezindabeni ngezifinyezo ezikhiqizwe yi-AI futhi bangenza izinqumo ezinolwazi ngaphandle kokudinga ukuhlola idatha eluhlaza. Ngokufanayo, ukuxhumana okusebenzayo (kubaphathi, izikhulu zokuthobela imithetho, njll.) kuyathuthuka lapho i-AI ilungiselela imibiko elula ukuyiqonda ngesimo sokuphepha kanye nezehlakalo (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zangempela). Lokhu akugcini nje ngokwakha ukuzethemba nokuhambisana ezindabeni zokuphepha ezingeni lobuholi kodwa futhi kusiza ukuthethelela ukutshalwa kwezimali kanye nezinguquko ngokuchaza ngokucacile izingozi kanye nezikhala ezitholwe yi-AI.

Ngokubambisana, lezi zinzuzo zisho ukuthi izinhlangano ezisebenzisa i-AI yokukhiqiza ekuphepheni kwe-cyber zingafinyelela isimo sokuphepha esiqinile ngezindleko zokusebenza ezingase ziphansi. Zingaphendula ezinsongweni ezazinzima ngaphambili, zimboze izikhala ezazingaqashelwa, futhi zithuthuke ngokuqhubekayo ngokusebenzisa i-feedback loops eqhutshwa yi-AI. Ekugcineni, i-AI yokukhiqiza inikeza ithuba lokudlula izitha ngokufanisa isivinini , ubukhulu, kanye nobuchwepheshe bokuhlasela kwesimanje ngokuzivikela okuyinkimbinkimbi ngokulinganayo. Njengoba inhlolovo eyodwa ithole, abaholi bebhizinisi nabama-cyber abangaphezu kwesigamu balindele ukutholakala kwezinsongo okusheshayo kanye nokunemba okwandisiwe ngokusebenzisa i-AI yokukhiqiza ([PDF] Global Cybersecurity Outlook 2025 | World Economic Forum) (I-AI Yokukhiqiza Ekuphepheni Kwe-Cyber: Ukubuyekezwa Okuphelele kwe-LLM ...) - ubufakazi bokuthi kunethemba ngezinzuzo zalobu buchwepheshe.

Izingozi Nezinselele Zokusebenzisa i-AI Ekhiqizayo Ekuphepheni Kwe-Cyber

Nakuba amathuba ebalulekile, kubalulekile ukubhekana ne-AI ekhiqizayo kwezokuphepha kwe-inthanethi ngamehlo avulekele izingozi nezinselele ezihilelekile. Ukuyethemba ngokungaqondile i-AI noma ukuyisebenzisa kabi kungaletha ubuthakathaka obusha. Ngezansi, sichaza ukukhathazeka okukhulu kanye nezingibe, kanye nomongo wento ngayinye:

• Ukusetshenziswa Kokuphikisana Yizigebengu Ze-Cyber: Amakhono afanayo okukhiqiza asiza abavikeli anganika amandla abahlaseli. Abenzi bokusongela sebevele besebenzisa i-AI yokukhiqiza ukwenza ama-imeyili obugebengu bokweba imininingwane ebucayi, ukudala abantu abangombombayi kanye namavidiyo ajulile obunjiniyela bezenhlalo, ukuthuthukisa i-malware ye-polymorphic eshintsha njalo ukuze igweme ukutholakala, futhi ize yenze ngokuzenzakalelayo izici zokugenca (Kuyini i-AI Ekhiqizayo Ku-Cybersecurity? - Amanethiwekhi e-Palo Alto). Cishe ingxenye (46%) yabaholi bezokuphepha kwe-cyber bakhathazekile ngokuthi i-AI yokukhiqiza izoholela ekuhlaselweni okuthuthukisiwe kakhulu kokuphikisana (Ukuphepha kwe-AI Yokukhiqiza: Amathrendi, Izinsongo kanye Namasu Okunciphisa). Lo "mjaho wezikhali ze-AI" usho ukuthi njengoba abavikeli besebenzisa i-AI, abahlaseli ngeke babe ngemuva kakhulu (eqinisweni, bangase babe phambili kwezinye izindawo, besebenzisa amathuluzi e-AI angalawulwa). Izinhlangano kumele zilungele izinsongo ezithuthukisiwe ze-AI ezivame kakhulu, eziyinkimbinkimbi, futhi okunzima ukuzilandela.

• Ukungaqondi kahle kanye nokungalungi kwe-AI: Amamodeli e-AI akhiqizayo angakhiqiza imiphumela enengqondo kodwa engalungile noma edukisayo - into eyaziwa ngokuthi ukungaboni kahle. Esimweni sokuphepha, i-AI ingase ihlaziye isigameko bese iphetha ngephutha ngokuthi ubuthakathaka obuthile buyimbangela, noma ingase ikhiqize iskripthi sokulungisa esinephutha esihluleka ukuqukatha ukuhlaselwa. Lawa maphutha angaba yingozi uma ethathwa ngokubonakalayo. Njengoba i-NTT Data ixwayisa, "i-AI ekhiqizayo ingase ikhiphe okuqukethwe okungelona iqiniso, futhi lesi simo sibizwa ngokuthi ukungaboni kahle... okwamanje kunzima ukukuqeda ngokuphelele" (Izingozi Zokuphepha ze-AI Ekhiqizayo kanye Nezindlela Zokumelana, kanye Nomthelela Wayo Ekuvikelekeni Kwe-Cyber ​​| I-NTT DATA Group). Ukuthembela ngokweqile ku-AI ngaphandle kokuqinisekiswa kungaholela emizamweni eqondiswe kabi noma umuzwa wamanga wokuphepha. Isibonelo, i-AI ingase ibeke uphawu ngamanga ohlelweni olubucayi njengoluphephile lapho lungalutholi, noma ngokuphambene nalokho, lubangele ukwethuka "ngokuthola" ukwephulwa okungakaze kwenzeke. Ukuqinisekiswa okuqinile kwemiphumela ye-AI kanye nokuba nabantu abasesimweni sezinqumo ezibalulekile kubalulekile ukunciphisa le ngozi.

• Okuhle Nokubi Okungamanga: Okuhlobene nokubona izinto ezingekho, uma imodeli ye-AI ingaqeqeshwanga kahle noma ilungiselelwe kabi, ingase ibike ngokweqile imisebenzi engemihle njengenonya (okuhle okungamanga) noma, okubi nakakhulu, iphuthelwe izinsongo zangempela (okungalungile okungamanga) (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity). Izexwayiso ezingamanga eziningi zinganqoba amaqembu okuphepha futhi ziholele ekukhathaleni okuqaphile (ukuhlehlisa izinzuzo zokusebenza kahle ze-AI ezithenjisiwe), kuyilapho ukutholwa okungafanele kushiya inhlangano iveziwe. Ukulungisa amamodeli okukhiqiza ukuze kube nebhalansi efanele kuyinselele. Indawo ngayinye ihlukile, futhi i-AI ingase ingasebenzi ngokushesha ngendlela efanele ngaphandle kwebhokisi. Ukufunda okuqhubekayo kuyinkemba esika nhlangothi zombili - uma i-AI ifunda ngempendulo egobile noma endaweni eshintshayo, ukunemba kwayo kungaguquguquka. Amaqembu okuphepha kumele aqaphe ukusebenza kwe-AI futhi alungise imingcele noma anikeze impendulo yokulungisa kumamodeli. Ezimweni ezibaluleke kakhulu (njengokutholwa kokungena kwengqalasizinda ebalulekile), kungaba ukuhlakanipha ukusebenzisa iziphakamiso ze-AI ngokuhambisana nezinhlelo ezikhona isikhathi esithile, ukuqinisekisa ukuthi ziyahambisana futhi ziyahambisana kunokuba kube nokungqubuzana.

• Ubumfihlo Nokuvuza Kwedatha: Izinhlelo ze-AI ezikhiqizayo zivame ukudinga idatha eningi yokuqeqeshwa nokusebenza. Uma lawa mamodeli esuselwe efwini noma engagcinwanga kahle, kunengozi yokuthi ulwazi olubucayi lungavuza. Abasebenzisi bangase bafake idatha yomuntu siqu noma idatha yomuntu siqu ngengozi kusevisi ye-AI (cabanga ngokucela i-ChatGPT ukuthi ifingqe umbiko wesigameko oyimfihlo), nokuthi leyo datha ingaba yingxenye yolwazi lomodeli. Ngempela, ucwaningo lwamuva nje luthole ukuthi ama-55% okufakwayo kumathuluzi e-AI akhiqizayo aqukethe ulwazi olubucayi noma oluqondakalayo lomuntu siqu, okuphakamisa ukukhathazeka okukhulu mayelana nokuvuza kwedatha (Ukuphepha kwe-AI Okukhiqizayo: Amathrendi, Izinsongo Nokunciphisa Amasu). Ngaphezu kwalokho, uma i-AI iqeqeshwe ngedatha yangaphakathi futhi ibuzwa ngezindlela ezithile, ingase ikhiphe izingcezu zaleyo datha ebucayi komunye umuntu. Izinhlangano kumele zisebenzise izinqubomgomo eziqinile zokuphatha idatha (isb. ukusebenzisa izimo ze-AI ezisendaweni noma ezizimele zezinto ezibucayi) futhi zifundise abasebenzi ngokunganamathiseli ulwazi oluyimfihlo kumathuluzi e-AI omphakathi. Imithethonqubo yobumfihlo (GDPR, njll.) nayo iyasetshenziswa - ukusebenzisa idatha yomuntu siqu ukuqeqesha i-AI ngaphandle kwemvume efanele noma ukuvikelwa kungaphula imithetho.

• Ukuphepha Nokukhohlisa Kwemodeli: Amamodeli e-AI akhiqizayo ngokwawo angaba yizisulu. Izitha zingase zizame ukufaka ubuthi bemodeli, zinikeze idatha enonya noma edukisayo ngesikhathi sokuqeqeshwa noma sokuqeqeshwa kabusha ukuze i-AI ifunde amaphethini angalungile (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity). Isibonelo, umhlaseli angase afake ubuthi ngengozi idatha ye-Intel ukuze i-AI yehluleke ukuqaphela i-malware yomhlaseli njengononya. Elinye isu ukufaka ngokushesha noma ukukhohlisa okukhiphayo, lapho umhlaseli ethola indlela yokukhipha okokufaka ku-AI okubangela ukuthi iziphathe ngezindlela ezingahlosiwe - mhlawumbe ukungazinaki izithiyo zayo zokuphepha noma ukwembula ulwazi okungafanele (njengezikhuthazo zangaphakathi noma idatha). Ngaphezu kwalokho, kunengozi yokugwema imodeli: abahlaseli benza okokufaka okuklanyelwe ngqo ukukhohlisa i-AI. Sikubona lokhu ezibonelweni eziphikisanayo - idatha ephazamisekile kancane umuntu ayibona njengejwayelekile kodwa i-AI ayihlukanisi kahle. Ukuqinisekisa ukuthi uchungechunge lokuhlinzekwa kwe-AI luphephile (ubuqotho bedatha, ukulawulwa kokufinyelela kwemodeli, ukuhlolwa kokuqina kokuphikisana) kuyingxenye entsha kodwa edingekayo yokuphepha kwe-cyber lapho kusetshenziswa la mathuluzi (Iyini i-AI Ekhiqizayo Kukuphepha Kwe-Cyber? - Palo Alto Networks).

• Ukuzethemba Ngokweqile Nokuguguleka Kwamakhono: Kunengozi ethambile yokuthi izinhlangano zingase zithembele kakhulu ku-AI futhi zivumele amakhono abantu aphele. Uma abahlaziyi abancane bethemba imiphumela ye-AI ngokungaboni, bangase bangathuthuki ukucabanga okubucayi kanye nokuqonda okudingekayo lapho i-AI ingatholakali noma ingalungile. Isimo okufanele sigwenywe yithimba lezokuphepha elinamathuluzi amahle kodwa angazi ukuthi lizosebenza kanjani uma lawo mathuluzi ewa (njengoba abashayeli bezindiza bethembele kakhulu ku-autopilot). Ukuzivocavoca njalo ngaphandle kosizo lwe-AI nokukhuthaza isimo sengqondo sokuthi i-AI ingumsizi, hhayi isazi esingenaphutha, kubalulekile ukugcina abahlaziyi abangabantu bebukhali. Abantu kumele bahlale bengabenzi bezinqumo abaphezulu, ikakhulukazi ngezinqumo ezithinta kakhulu.

• Izinselele Zokuziphatha Nokuthobela Imithetho: Ukusetshenziswa kwe-AI ekuphepheni kwe-cyber kuphakamisa imibuzo yokuziphatha futhi kungabangela izinkinga zokuthobela imithetho. Isibonelo, uma uhlelo lwe-AI lufaka isisebenzi kabi njengomuntu wangaphakathi ononya ngenxa yento engavamile, lungalimaza ngokungafanele idumela noma umsebenzi walowo muntu. Izinqumo ezenziwe yi-AI zingaba yinto engacacile (inkinga "yebhokisi elimnyama"), okwenza kube nzima ukuchaza kubahloli noma abalawuli ukuthi kungani kuthathwe izinyathelo ezithile. Njengoba okuqukethwe okukhiqizwe yi-AI kusabalala kakhulu, ukuqinisekisa ukucaca nokugcina ukuzibophezela kubalulekile. Abalawuli baqala ukuhlola i-AI - uMthetho we-AI we-EU, isibonelo, uzobeka izidingo ezinhlelweni ze-AI "ezinobungozi obukhulu", futhi i-AI yokuphepha kwe-cyber ingase iwele kuleso sigaba. Izinkampani kuzodingeka zihambe ngalezi zimiso futhi mhlawumbe zilandele amazinga afana ne-NIST AI Risk Management Framework ukuze zisebenzise i-AI ekhiqizayo ngokuzibophezela (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuphepheni Kwe-Cyber? Izibonelo Eziyi-10 Zangempela). Ukuthobela imithetho kudlulela nasekunikezeni ilayisense: ukusebenzisa amamodeli omthombo ovulekile noma ezinkampani zangaphandle kungaba nemigomo evimbela ukusetshenziswa okuthile noma edinga ukuthuthukiswa kokwabelana.

Ngamafuphi, i-AI yokukhiqiza akuyona into elula - uma ingasetshenziswa ngokucophelela, ingaletha ubuthakathaka obusha ngisho noma ixazulula obunye. Ucwaningo lwe-World Economic Forum lwango-2024 luqokomise ukuthi cishe ama-47% ezinhlangano zikhomba intuthuko ku-AI yokukhiqiza ngabahlaseli njengento ekhathazayo eyinhloko, okwenza kube "umthelela obaluleke kakhulu we-AI yokukhiqiza" ekuphepheni kwe-cyber ([PDF] Global Cybersecurity Outlook 2025 | World Economic Forum) (I-AI Yokukhiqiza Ekuphepheni Kwe-Cyber: Ukubuyekezwa Okuphelele kwe-LLM ...). Ngakho-ke izinhlangano kumele zamukele indlela elinganiselayo: zisebenzisa izinzuzo ze-AI ngenkathi zilawula ngokuqinile lezi zingozi ngokubusa, ukuhlola, kanye nokuqapha kwabantu. Okulandelayo sizoxoxa ngendlela yokufinyelela leyo bhalansi ngokoqobo.

Umbono Wesikhathi Esizayo: Indima Eguqukayo ye-AI Ekhiqizayo Ekuphepheni Kwe-Cyber

Uma sibheka phambili, i-AI ekhiqizayo isilungele ukuba yingxenye ebalulekile yesu lokuphepha kwe-cyber - futhi ngokufanayo, ithuluzi abaphikisi be-cyber abazoqhubeka nokulisebenzisa. Ukuguquguquka kwe-cat-and-mouse kuzoshesha, nge-AI kuzo zombili izinhlangothi zocingo. Nazi ezinye zezindlela zokubheka phambili zokuthi i-AI ekhiqizayo ingabumba kanjani ukuphepha kwe-cyber eminyakeni ezayo:

• Ukuvikelwa Kwe-Cyber ​​Okukhuliswe Nge-AI Kuba Yindinganiso: Ngo-2025 nangale kwalokho, singalindela ukuthi izinhlangano eziningi eziphakathi kuya kwezinkulu zizobe sezifake amathuluzi aqhutshwa yi-AI emisebenzini yazo yokuphepha. Njengoba nje ama-antivirus nama-firewall ejwayelekile namuhla, ama-copilot e-AI kanye nezinhlelo zokuthola okungavamile zingaba yizingxenye eziyisisekelo zezakhiwo zokuphepha. Lawa mathuluzi cishe azoba akhethekile kakhulu - isibonelo, amamodeli ahlukile e-AI ahlelwe kahle ngokuphepha kwamafu, ukuqapha idivayisi ye-IoT, ukuphepha kwekhodi yohlelo lokusebenza, njalo njalo, konke kusebenza ngokubambisana. Njengoba esinye isibikezelo siphawula, "ngo-2025, i-AI ekhiqizayo izoba yingxenye yokuphepha kwe-cyber, ivumele izinhlangano ukuthi zivikele ezinsongweni eziyinkimbinkimbi nezishintshayo ngendlela esebenzayo" (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ku-Cybersecurity). I-AI izothuthukisa ukutholwa kwezinsongo ngesikhathi sangempela, yenze ngokuzenzakalelayo izenzo eziningi zokuphendula, futhi isize amaqembu okuphepha ukuphatha inani elikhulu kakhulu ledatha kunaleyo abengayenza ngesandla.

• Ukufunda Okuqhubekayo Nokuzivumelanisa Nezimo: Izinhlelo ze-AI ezikhiqizayo zesikhathi esizayo ku-cyber zizoba ngcono ekufundeni ngokushesha ezigamekweni ezintsha kanye nobuhlakani bezinsongo, zivuselele ulwazi lwazo ngesikhathi sangempela. Lokhu kungaholela ekuzivikeleni okuguquguqukayo ngempela - cabanga nge-AI efunda ngomkhankaso omusha we-phishing ohlasela enye inkampani ekuseni futhi ntambama isivele ilungise izihlungi ze-imeyili zenkampani yakho ngokuphendula. Izinsizakalo zokuphepha ze-AI ezisekelwe efwini zingenza kube lula lolu hlobo lokufunda ngokuhlanganyela, lapho ukuqonda okungaziwa okuvela enhlanganweni eyodwa kuzuzisa bonke ababhalisile (okufana nokwabelana nge-intel ngezinsongo, kodwa okuzenzakalelayo). Kodwa-ke, lokhu kuzodinga ukuphathwa ngokucophelela ukuze kugwenywe ukwabelana ngolwazi olubucayi nokuvimbela abahlaseli ukuthi bangafaka idatha embi kumamodeli abiwe.

• Ukuhlangana kwe-AI kanye ne-Cybersecurity Talent: Iqoqo lamakhono ochwepheshe be-cybersecurity lizothuthuka ukuze lifake ubuchwepheshe ku-AI kanye nesayensi yedatha. Njengoba nje abahlaziyi banamuhla befunda izilimi zemibuzo kanye nokubhala, abahlaziyi bakusasa bangase bahlele njalo amamodeli e-AI noma babhale “izincwadi zokudlala” ze-AI ukuze zisebenze. Singase sibone izindima ezintsha ezifana “noMqeqeshi Wezokuphepha we-AI” noma “uNjiniyela we-AI we-Cybersecurity” – abantu abangochwepheshe ekuvumelaniseni amathuluzi e-AI nezidingo zenhlangano, ukuqinisekisa ukusebenza kwawo, nokuqinisekisa ukuthi asebenza ngokuphephile. Ngakolunye uhlangothi, izinto ezicatshangelwayo zokuphepha kwe-cyber zizothinta kakhulu ukuthuthukiswa kwe-AI. Izinhlelo ze-AI zizokwakhiwa ngezici zokuphepha kusukela phansi kuya phezulu (ukwakheka okuphephile, ukutholwa kokuphazamiseka, amalogi okuhlola ezinqumweni ze-AI, njll.), kanye nezinhlaka ze- AI ethembekile (efanelekile, echazwe kahle, eqinile, futhi evikelekile) zizoqondisa ukuthunyelwa kwazo ezimweni ezibalulekile zokuphepha.

• Ukuhlaselwa Okunamandla E-AI Okuyinkimbinkimbi: Ngeshwa, isimo sosongo sizophinde sithuthuke nge-AI. Silindele ukusetshenziswa kaningi kwe-AI ukuthola ubuthakathaka obungenaso usuku, ukwakha ubugebengu bokweba imininingwane ebucayi obuqondiswe kakhulu (isb. i-AI isusa imithombo yezokuxhumana ukuze idale isicupho esilungiselelwe kahle), kanye nokukhiqiza amazwi noma amavidiyo akhohlisayo ajulile ukuze kugwenywe ukuqinisekiswa kwe-biometric noma ukwenza ukukhwabanisa. Ama-ejenti okugenca azenzakalelayo angavela angenza ngokuzimela ukuhlaselwa kwezigaba eziningi (ukuhlola, ukuxhaphaza, ukunyakaza okuseceleni, njll.) ngaphandle kokuqondiswa komuntu okuncane. Lokhu kuzocindezela abavikeli ukuthi bathembele ku-AI - empeleni ukwenza okuzenzakalelayo vs. ukwenza okuzenzakalelayo. Ezinye izihlaselo zingase zenzeke ngesivinini somshini, njenge-AI bots ezama izinkulungwane ze-imeyili zokweba imininingwane eyimfihlo ukuze zibone ukuthi iyiphi edlula izihlungi. Ukuzivikela kwe-Cyber ​​kuzodingeka kusebenze ngesivinini esifanayo kanye nokuguquguquka ukuze kuhambisane (Kuyini i-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? - Amanethiwekhi e-Palo Alto).

• Umthethonqubo kanye ne-AI yokuziphatha kahle ekuphepheni: Njengoba i-AI igxila kakhulu emisebenzini yokuphepha kwe-inthanethi, kuzoba nokuhlolwa okukhulu futhi mhlawumbe nomthethonqubo ukuqinisekisa ukuthi lezi zinhlelo ze-AI zisetshenziswa ngendlela efanele. Singalindela izinhlaka nezindinganiso eziqondene ne-AI ekuphepheni. Ohulumeni bangase babeke iziqondiso zokucaca - isib., ezidinga ukuthi izinqumo ezibalulekile zokuphepha (njengokuqeda ukufinyelela kwesisebenzi emisebenzini enobungozi esolwayo) zingenziwa yi-AI yodwa ngaphandle kokubuyekezwa ngabantu. Kungase kube nezitifiketi zemikhiqizo yokuphepha ye-AI, ukuqinisekisa abathengi ukuthi i-AI ihlolwe ukucwasa, ukuqina, kanye nokuphepha. Ngaphezu kwalokho, ukubambisana kwamazwe ngamazwe kungakhula eduze kwezinsongo ze-cyber ezihlobene ne-AI; isibonelo, izivumelwano zokuphatha ulwazi olungelulo oludalwe yi-AI noma izimiso ngokumelene nezikhali ezithile ze-cyber eziqhutshwa yi-AI.

• Ukuhlanganiswa ne-Broad AI kanye ne-IT Ecosystems: I-AI ekhiqizayo ekuphepheni kwe-cyber cishe izohlanganiswa nezinye izinhlelo ze-AI kanye namathuluzi okuphatha i-IT. Isibonelo, i-AI ephatha ukulungiswa kwenethiwekhi ingasebenza ne-AI yokuphepha ukuqinisekisa ukuthi izinguquko azivuli izikhala. Ukuhlaziywa kwebhizinisi okuqhutshwa yi-AI kungase kwabelane ngedatha nama-AI okuphepha ukuze kuhlotshaniswe ukungalingani (njengokwehla okungazelelwe kokuthengisa ngenkinga yewebhusayithi engaba khona ngenxa yokuhlaselwa). Empeleni, i-AI ngeke ihlale endaweni eyodwa - izoba yingxenye yendwangu ehlakaniphile enkulu yokusebenza kwenhlangano. Lokhu kuvula amathuba okuphathwa kwezingozi okuphelele lapho idatha yokusebenza, idatha yosongo, kanye nedatha yokuphepha ebonakalayo kungahlanganiswa yi-AI ukuze kunikezwe umbono wama-degree angu-360 ngesimo sokuphepha kwenhlangano.

Ngokuhamba kwesikhathi, ithemba liwukuthi i-AI ekhiqizayo izosiza ekulinganiseni ibhalansi ukuze isekele abavikeli. Ngokusingatha ubukhulu nobunzima bezindawo ze-IT zanamuhla, i-AI ingenza i-cyberspace ibe nokuzivikela kakhudlwana. Kodwa-ke, luhambo, futhi kuzoba nobunzima obukhulayo njengoba sithuthukisa lobu buchwepheshe futhi sifunda ukubethemba ngendlela efanele. Izinhlangano ezihlala zinolwazi futhi zitshala imali ekwamukelweni kwe -AI okunesibopho kwezokuphepha cishe yizo ezisesimweni esihle kakhulu sokubhekana nezinsongo zesikhathi esizayo.

Njengoba umbiko wakamuva wezindlela zokuphepha kwe-inthanethi kaGartner uphawulile, "ukuvela kwamacala okusetshenziswa kwe-AI ekhiqizayo (kanye nezingozi) kudala ingcindezi yokuguqulwa" (Izindlela Zokuphepha Kwe-inthanethi: Ukuqina Ngokuguqulwa - Gartner). Labo abavumelana nezimo bazosebenzisa i-AI njengomngani onamandla; labo abasalela emuva bangase bazithole bedlulwa yizitha ezinikwe amandla yi-AI. Iminyaka embalwa ezayo izoba yisikhathi esibalulekile ekuchazeni ukuthi i-AI iyibumba kanjani kabusha inkundla yempi ye-inthanethi.

Izinto Eziwusizo Zokusebenzisa i-AI Ekhiqizayo Kwezokuphepha Kwe-Cyber

Kumabhizinisi ahlola ukuthi angayisebenzisa kanjani i-AI ekhiqizayo esu lawo lokuphepha kwe-inthanethi, nansi eminye imibono nezincomo ezisebenzayo zokuqondisa ukwamukelwa okunesibopho nokuphumelelayo:

1. Qala ngeMfundo Nokuqeqeshwa: Qinisekisa ukuthi ithimba lakho lezokuphepha (kanye nabasebenzi be-IT ababanzi) bayaqonda ukuthi yini i-AI ekhiqizayo engayenza nengayikwenzi. Nikeza ukuqeqeshwa ngezisekelo zamathuluzi okuphepha aqhutshwa yi-AI futhi ubuyekeze izinhlelo zakho zokuqwashisa ngokuphepha kubo bonke abasebenzi ukuze babhekane nezinsongo ezisebenzisa i-AI. Isibonelo, fundisa abasebenzi ukuthi i-AI ingakhiqiza kanjani ukukhwabanisa kobugebengu bokweba imininingwane ebucayi kanye nezingcingo ezijulile. Ngesikhathi esifanayo, qeqesha abasebenzi ngokusetshenziswa okuphephile nokuvunyiwe kwamathuluzi e-AI emsebenzini wabo. Abasebenzisi abanolwazi oluhle banamathuba amancane okuphatha kabi i-AI noma babe yizisulu zokuhlaselwa okuthuthukisiwe yi-AI (Ingasetshenziswa Kanjani I-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela).

2. Chaza Izinqubomgomo Zokusebenzisa i-AI Ezicacile: Phatha i-AI ekhiqizayo njenganoma yiluphi ubuchwepheshe obunamandla - ngokuphatha. Yakha izinqubomgomo ezichaza ukuthi ubani ongasebenzisa amathuluzi e-AI, ukuthi yimaphi amathuluzi avunyelwe, nokuthi ngaziphi izinjongo. Faka iziqondiso zokuphatha idatha ebucayi (isb. ukungafakwa kwedatha eyimfihlo kumasevisi e-AI angaphandle) ukuvimbela ukuvuza. Isibonelo, ungavumela amalungu eqembu lezokuphepha kuphela ukuthi asebenzise umsizi we-AI wangaphakathi ukuze aphendule ezigamekweni, futhi ukumaketha kungasebenzisa i-AI ehlolwe ngokuqukethwe - wonke umuntu uvinjelwe. Izinhlangano eziningi manje zikhuluma ngokusobala nge-AI ekhiqizayo ezinqubomgomo zazo ze-IT, futhi izinhlangano eziholayo zezindinganiso zikhuthaza izinqubomgomo zokusetshenziswa okuphephile kunokuvinjelwa okuqondile (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela). Qiniseka ukuthi udlulisela le mithetho kanye nesizathu sayo kubo bonke abasebenzi.

3. Nciphisa "i-Shadow AI" kanye Nokusetshenziswa Kokuqapha: Ngokufana ne-shadow IT, "i-shadow AI" ivela lapho abasebenzi beqala ukusebenzisa amathuluzi noma izinsizakalo ze-AI ngaphandle kolwazi lwe-IT (isb. unjiniyela osebenzisa umsizi wekhodi ye-AI ongagunyaziwe). Lokhu kungaletha izingozi ezingabonakali. Sebenzisa izinyathelo zokuthola nokulawula ukusetshenziswa kwe-AI okungagunyaziwe. Ukuqapha inethiwekhi kungamaka ukuxhumana nama-API e-AI adumile, futhi izinhlolovo noma ukuhlolwa kwamathuluzi kungaveza ukuthi abasebenzi basebenzisa ini. Nikeza ezinye izindlela ezivunyiwe ukuze abasebenzi abanezinhloso ezinhle bangalingeki ukuba ngamasela (isibonelo, hlinzeka nge-akhawunti esemthethweni ye-ChatGPT Enterprise uma abantu beyithola iwusizo). Ngokuletha ukusetshenziswa kwe-AI ekukhanyeni, amaqembu okuphepha angahlola futhi aphathe ingozi. Ukuqapha nakho kubalulekile - bhala phansi imisebenzi yamathuluzi e-AI kanye nemiphumela ngangokunokwenzeka, ukuze kube nomzila wokuhlola ngezinqumo ezithonywe yi-AI (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zezwe Langempela).

4. Sebenzisa i-AI Ngokuzivikela – Ungasali Ngemuva: Qaphela ukuthi abahlaseli bazosebenzisa i-AI, ngakho-ke nokuzivikela kwakho kufanele. Khomba izindawo ezimbalwa ezinomthelela omkhulu lapho i-AI ekhiqizayo ingasiza khona ngokushesha imisebenzi yakho yokuphepha (mhlawumbe ukuqapha ukuhlaselwa, noma ukuhlaziywa kwamalogi okuzenzakalelayo) bese usebenzisa amaphrojekthi okuhlola. Khulisa ukuzivikela kwakho ngesivinini kanye nesilinganiso se-AI ukuze ulwe nezinsongo ezihamba ngokushesha (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ku-Cybersecurity? Izibonelo Eziyi-10 Zezwe Langempela). Ngisho nokuhlanganiswa okulula, njengokusebenzisa i-AI ukufingqa imibiko ye-malware noma ukukhiqiza imibuzo yokuzingela izinsongo, kungonga amahora abahlaziyi. Qala kancane, hlola imiphumela, futhi uphindaphinde. Impumelelo izokwakha icala lokwamukelwa kwe-AI ebanzi. Umgomo ukusebenzisa i-AI njengesiphindaphindi samandla – isibonelo, uma ukuhlaselwa kwe-phishing kudlula ideski lakho losizo, faka i-classifier ye-imeyili ye-AI ukuze unciphise lelo volumu ngokushesha.

5. Tshala imali emikhubeni ye-AI evikelekile nenezimiso zokuziphatha: Uma usebenzisa i-AI ekhiqizayo, landela imikhuba yokuthuthukisa nokusabalalisa evikelekile. Sebenzisa amamodeli azimele noma azimele emisebenzini ebucayi ukuze ugcine ukulawula idatha. Uma usebenzisa izinsizakalo ze-AI zenkampani yangaphandle, buyekeza izindlela zabo zokuphepha nobumfihlo (ukubethela, izinqubomgomo zokugcina idatha, njll.). Faka izinhlaka zokuphathwa kwengozi ye-AI (njenge-NIST's AI Risk Management Framework noma isiqondiso se-ISO/IEC) ukuze ubhekane nezinto ezinjengokubandlulula, ukuchaza, kanye nokuqina kumathuluzi akho e-AI (Ingasetshenziswa Kanjani i-AI Ekhiqizayo Ekuvikelekeni Kwe-Cyber? Izibonelo Eziyi-10 Zangempela). Hlela futhi izibuyekezo/ama-patches amamodeli njengengxenye yokulungisa - amamodeli e-AI angaba "nobuthakathaka" nawo (isb. angadinga ukuqeqeshwa kabusha uma eqala ukukhukhuleka noma uma kutholakala uhlobo olusha lokuhlasela okuphikisayo kumodeli). Ngokubhaka ukuphepha nokuziphatha ekusetshenzisweni kwakho kwe-AI, wakha ukwethembana emiphumeleni futhi uqinisekise ukuhambisana nemithetho emisha.

6. Gcina Abantu Besendaweni Ezungezile: Sebenzisa i-AI ukusiza, hhayi ukufaka esikhundleni ngokuphelele, isahlulelo somuntu ekuphepheni kwe-cyber. Nquma amaphuzu ezinqumo lapho kudingeka khona ukuqinisekiswa komuntu (isibonelo, i-AI ingase ibhale umbiko wesigameko, kodwa umhlaziyi uyibukeza ngaphambi kokusatshalaliswa; noma i-AI ingase iphakamise ukuvimba i-akhawunti yomsebenzisi, kodwa umuntu uyayivuma leso senzo). Lokhu akugcini nje ngokuvimbela amaphutha e-AI ukuthi angahlolwa, kodwa futhi kusiza ithimba lakho ukuthi lifunde ku-AI futhi okuphambene nalokho. Khuthaza ukuhamba komsebenzi ngokubambisana: abahlaziyi kufanele bazizwe bekhululekile ukungabaza imiphumela ye-AI nokwenza ukuhlolwa kokuqonda. Ngokuhamba kwesikhathi, le ngxoxo ingathuthukisa kokubili i-AI (ngokuphendula) kanye namakhono abahlaziyi. Empeleni, klama izinqubo zakho ngendlela yokuthi i-AI namandla abantu ahambisane - i-AI iphatha ivolumu kanye nesivinini, abantu baphatha ukungaqondakali kanye nezinqumo zokugcina.

7. Kala, Gada, Futhi Ulungise: Okokugcina, phatha amathuluzi akho e-AI akhiqizayo njengezingxenye eziphilayo zemvelo yakho yokuphepha. Kala njalo ukusebenza kwawo - ingabe anciphisa izikhathi zokuphendula ezigamekweni? Ubamba izinsongo kusenesikhathi? Izinga le-positive elingamanga lihamba kanjani? Cela impendulo eqenjini: ingabe izincomo ze-AI ziwusizo, noma ingabe zidala umsindo? Sebenzisa lezi zilinganiso ukuze uthuthukise amamodeli, ubuyekeze idatha yokuqeqesha, noma ulungise indlela i-AI ehlanganiswe ngayo. Izinsongo ze-cyber nezidingo zebhizinisi ziyashintsha, futhi amamodeli akho e-AI kufanele avuselelwe noma aqeqeshwe kabusha ngezikhathi ezithile ukuze ahlale esebenza kahle. Yiba nohlelo lokuphathwa kwamamodeli, okuhlanganisa ukuthi ubani obhekene nokugcinwa kwawo nokuthi ubuyekezwa kangaki. Ngokuphatha ngenkuthalo umjikelezo wokuphila kwe-AI, uqinisekisa ukuthi ihlala iyimpahla, hhayi umthwalo wemfanelo.

Ekuphetheni, i-AI ekhiqizayo ingathuthukisa kakhulu amakhono okuphepha kwe-cyber, kodwa ukwamukelwa ngempumelelo kudinga ukuhlela okucatshangelwe kahle kanye nokuqapha okuqhubekayo. Amabhizinisi afundisa abantu bawo, abeke iziqondiso ezicacile, futhi ahlanganise i-AI ngendlela elinganiselayo nephephile azovuna imivuzo yokuphathwa kwezinsongo okusheshayo nokuhlakaniphile. Lezo zifundo zinikeza imephu yomgwaqo: hlanganisa ubuchwepheshe babantu kanye nokwenza ngokuzenzakalela kwe-AI, zimboze izisekelo zokubusa, futhi zilondoloze ukuguquguquka njengoba ubuchwepheshe be-AI kanye nesimo sezinsongo kuqhubeka ngokungenakugwenywa.

Ngokuthatha lezi zinyathelo ezisebenzayo, izinhlangano zingaphendula ngokuzethemba umbuzo othi “I-AI ekhiqizayo ingasetshenziswa kanjani ekuphepheni kwe-cyber?” – hhayi nje ngokombono, kodwa nasekusebenzeni kwansuku zonke – ngaleyo ndlela ziqinise ukuzivikela kwazo ezweni lethu eliya ngokuya likhula ngokwedijithali kanye neliqhutshwa yi-AI. (I-AI Ekhiqizayo Ingasetshenziswa Kanjani Ekuphepheni Kwe-Cyber) 

Amaphepha Amhlophe ongase uthande ukuwafunda ngemva kwalokhu:

🔗 Imisebenzi Engeke Ishintshwe yi-AI kanye Nemisebenzi Ezothathelwa Indawo yi-AI?
Hlola umbono womhlaba wonke wokuthi yiziphi izindima eziphephile ezenzakalelayo nokuthi yiziphi eziphephile.

🔗 Ingabe i-AI ingakwazi ukubikezela imakethe yamasheya?
Ukubheka ngokucophelela imikhawulo, impumelelo, kanye nezinganekwane mayelana nekhono le-AI lokubikezela ukunyakaza kwemakethe.

🔗 Yini i-AI ekhiqizayo engathembelwa kuyo ngaphandle kokungenelela komuntu?
Qonda ukuthi i-AI ingasebenza kuphi ngokuzimela nokuthi ukuqondiswa komuntu kusabalulekile kuphi.