Zimbalwa izinsuku ezedlule iqembu labacwaningi kusuka emanyuvesi asePadua (Italy) naseDelft (Netherlands) wazise ngokushicilela ulwazi mayelana indlela yokusebenzisa umshini wokufunda ukuze udale kabusha i-PIN khodi kufakwe kuvidiyo eqoshiwe yendawo yokungena ku-ATM embozwe ngesandla.
Uma ufaka i-PIN enamadijithi angu-4, Amathuba okubikezela ikhodi efanele alinganiselwa ku-41%, kunikezwe ithuba lokwenza imizamo emithathu ngaphambi kokuvimba. Kumakhodi ephinikhodi anezinhlamvu ezi-5, amathuba okuqagela abe ngu-30%.
Ukwengeza, okunye ukuhlola kwenziwa lapho amavolontiya angu-78 azama ukubikezela ikhodi ye-PIN kumavidiyo afanayo arekhodiwe. Kulokhu, amathuba okubikezela okuyimpumelelo ayengu-7,92% ngemizamo emithathu.
Encazelweni yendlela esetshenzisiwe, kushiwo lokho Lapho iphaneli yedijithali ye-ATM imbozwe ngentende yesandla, ingxenye yesandla esifakwayo ihlala ingamboziwe, yini kwanele ukubikezela ukuchofoza ukushintsha indawo yesandla kanye nokugudluzwa kweminwe okungavalwanga ngokuphelele.
Ama-ATM amelela asetshenziswa kakhulu ohlelweni lokukhipha imali. I-European Central Bank ibike ukuhoxiswa kwemali okungaphezu kwezigidigidi eziyi-11 kanye nokulayishwa / ukulanda okwenziwe kuma-ATM aseYurophu ngo-2019.
Nakuba ama-ATM enziwe izinguquko ezihlukahlukene zobuchwepheshe, izinombolo zokukhomba umuntu (ama-PIN) kuseyizindlela zokuqinisekisa ezivame kakhulu zala madivayisi.Ngeshwa, indlela ye-PIN isengcupheni yokuhlaselwa okwenziwa ngamakhamera afihliwe afakwe eduze kwe-ATM ukuze acuphe ikhiphedi.
Lapho uhlaziya okokufaka kwedijithi ngayinye, isistimu ayibandakanyi okhiye abangakwazi ukucindezelwa, kucatshangelwa ukuma kwesandla esisibekeleyo, futhi iphinda ibale ukwahluka okungenzeka kakhulu kwengcindezi ngokusekelwe endaweni yesandla sokucindezela, ngokuhlobene nendawo yokhiye. Ukwandisa amathuba okuthola okokufaka, umsindo wokuchofoza ungaphinda urekhodwe, ohluke kancane kukhiye ngamunye.
Ukuhlolwa kusebenzise isistimu yokufunda yomshini ngokusekelwe ekusetshenzisweni kwenethiwekhi ye-convolutional neural (CNN) kanye nenethiwekhi ye-neural eqhubekayo esekelwe ku-LSTM (Inkumbulo Yesikhathi Esifushane Eside) izakhiwo. I-CNN yayinesibopho sokukhipha idatha yendawo yohlaka ngalunye, futhi i-LSTM yasebenzisa le datha ukuze ikhiphe amaphethini ahluka ngokuhamba kwesikhathi. Imodeli yaqeqeshwa ekurekhodweni kwevidiyo yokufaka ikhodi ye-PIN ngabantu abahlukene abangu-58 besebenzisa izindlela zokungenela ezikhethwe abahlanganyeli (umhlanganyeli ngamunye ufake amakhodi ahlukene ayi-100, okungukuthi, izibonelo zokungenela ezingama-5800 zisetshenziswe ekuqeqeshweni). Phakathi nokuqeqeshwa, kwavezwa ukuthi abasebenzisi abaningi basebenzisa enye yezindlela ezintathu eziyinhloko zokufihla ukungena.
Ukuze uqeqeshe imodeli yokufunda yomshini, iseva esekelwe kuphrosesa ye-Xeon E5-2670 eno-128 GB we-RAM namakhadi e-Tesla K20m amathathu anememori engu-5 GB ngalinye. Ingxenye yesoftware ibhalwe ngePython kusetshenziswa umtapo wezincwadi wakwaKeras kanye nenkundla yeTensorflow. Njengoba amaphaneli okungena e-ATM ehlukile futhi umphumela wokubikezela uncike ezicini ezifana nosayizi wokhiye ne-topology, ukuqeqeshwa okuhlukene kuyadingeka ohlotsheni ngalunye lwephaneli.
Njengesilinganiso sokuzivikela ngokumelene nendlela yokuhlasela ehlongozwayo, Kunconywa ukusebenzisa amakhodi e-PIN anezinhlamvu ezi-5 esikhundleni sika-4 uma kwenzeka, futhi uzame ukuvala isikhala esiningi sokungena ngesandla sakho (Indlela isasebenza uma cishe u-75% wendawo yokungena ivalwe ngesandla). Kunconywa ukuthi abakhiqizi be-ATM basebenzise izikrini ezikhethekile zokuzivikela ezifihla umnyango, kanye namaphaneli okungena okungewona ama-mechanical, kodwa athintekayo, indawo yezinombolo lapho ishintsha khona ngokungahleliwe.
Ekugcineni, uma unentshisekelo yokwazi kabanzi ngakho, ungaxhumana nemininingwane Kulesi sixhumanisi esilandelayo.