← Komawa blog
Counter-Forensics An Bayyana: Amfani da AI don Gano Canjin Hoto da Karya

Counter-Forensics An Bayyana: Amfani da AI don Gano Canjin Hoto da Karya

Kowane hoto na bayar da wata labari. Amma a wannan filin yaƙi na dijital na yanzu, wannan labarin na iya sake rubutawa.

Masu aikawa da yaudara na anti-forensic suna cire shaidu, yanke wuraren alamu, har ma suna amfani da cike bisa ga abin ciki (content-aware fill) don share duk wata alamar inda ko lokacin da hoton ya dauke.

Ga masu bincike masu zaman kansu, hukumomin tsaro, da ƙwararrun OSINT, waɗannan ci gaba suna gabatar da ƙalubale mai ƙaruwa.

Ta yaya zaku kasance mataki na gaba idan shaidar kanta ana gyara ta?

Amsar ta ta'allaka ne a rungumar matakan kariya da aka ƙafa ta AI. A cikin wannan labarin, za mu zurfafa cikin mafi ci gaba na dabarun karya hoto da za ku fuskanta, mu bayyana kayan aikin AI na zamani da aka tsara don gano yaudara, kuma mu raba mafi kyawun hanyoyin haɗawa waɗannan ƙwarewar cikin tsarin bincikenku. A ƙarshe, za ku sami taswira mai kyau don bayyana maƙallan hoto mafi rikitarwa.

Fahimtar Kundin Dabarun Counter-Forensics

Kafin mu xi hangen hanyar gano, bari mu bayyana dabarun da masu yaudara na anti-forensic suke dogaro da su:

  • Cire Hayar Hayaniya na Kamara (Sensor Noise Removal): Kamara ta barin siffar hayaniya mai laushi wadda take kama da sawun yatsa. Cire ko murƙushe waɗannan alamomin zai iya ɓoye tushen na'urar da hoton ya fito.
  • Yankewa da Tsara Wuri (Landmark Cropping and Framing): Yanke hoto cikin dabara yana iya cire alamomin geolocation kamar tutoci titin, fuskokin gine-gine, ko siffofin yanayi.
  • Cike Bisa ga Abun Ciki (Content-Aware Fill): Kayan gyara na zamani na iya maye gurbin abubuwa marasa so ko bango ba tare da barin alamar gyara ba, suna haɗa laushi da haske don ɓoye gyara.
  • Deepfakes da Samfuran Kirkira (Deepfakes and Generative Models): Fuskar da aka samar da AI ko yanayi na hotuna na iya saka mutane ko muhallin da bai taɓa wanzu ba.
  • Cire Metadata (Metadata Scrubbing): Cire bayanan EXIF yana cire timestamps, GPS coordinates, da bayani game da samfurin kamara, yana rage shaidu.

Waɗannan hanyoyin suna haɓowa cikin sauri. Abin da ya yi a baya ta duban ido ya sauya zuwa buƙatar bincike ta hanyar algorithms.

Kayan Aikin AI don Warware Yaudara

AI na motsa fili wajen bayar da damar gano karya hoto a girma. A ƙasa akwai muhimman dabarun da ake amfani da su ta AI da ke sauya tabbaci hoto.

  1. Nazarin Tsarin Hayaniya (Noise Pattern Analysis): Algorithms masu ci gaba na iya sake gina ko kwatanta siginan hayaniya na na'ura (sensor noise), wanda ake kira photo response non-uniformity (PRNU). Ta hanyar daidaita hayaniya saura a hoto da ake zargi da shi da kundin samfurin kamara da aka sani, za ku iya tabbatar ko hoton ya fito daga na'urar da ake zargin.
  2. Netwot ɗin Gano Deepfake (Deepfake Detection Networks): Neural networks da aka horar a kan dubban hotuna na hakika da waɗanda AI ta ƙirƙira na iya gano alamomin ƙirƙira. Daga yanayin ko shafe ido maras dabi’a zuwa rashin daidaito a fuska, waɗannan na’urorin gano suna haskaka wuraren da za a iya samun canji.
  3. Nazarin Matakin Kuskure tare da Haɓaka AI (Error Level Analysis with AI Enhancement): Nazarin matakin kuskure na gargajiya yana nuna yankuna daban-daban na ingancin lambar matsawa. sigogi na AI sun tsabtace wannan tunani, suna amfani da convolutional neural networks don bambanta alamomin JPEG na gaskiya daga waɗanda aka ƙara ta gyare-gyare cikin zaɓi.
  4. Masu Gano Rashin Daidaito na Ma’anoni (Semantic Inconsistency Scanners): Waɗannan samfuran suna kimanta daidaiton ma’anar yanayi. Shin hasken motar yana dacewa da inuwa a cikin gine-gine a baya? Shin ganye sun dace da yanayin yankin? AI scanners suna iya nuna rikice-rikice a hangen nesa, auna, da haske.
  5. Samfurori na Geolocation bisa ga Mahalli (Contextual Geolocation Models): Ko da EXIF data ya ɓace kuma alamomin wuri sun ɓuya, AI na iya nazarin alamomin muhalli. Salon gine-gine, nau'in tsiro, fonts na tutoci, da layukan shimfiɗa suna shigar da algorithms na geolocation waɗanda ke hasashen madaidaicin haɗin kai tare da ƙimar amincewa.
  6. Injinan Hasashen Lokaci (Temporal Inference Engines): Ta hanyar nazarin alamomin yanayi kamar dusar ƙanƙara, launin ganyen itace, ko kusurwar haske, waɗannan kayan aikin AI na kimanta lokacin shekara ko ranar da hoton ya dauka, yana fallasa hotuna da ake labarta su daga wata yanayi daban.
  7. Masu Gano Anomalies na Metadata: Ko da metadata na iya ɓacewa, AI na duba headers na fayil, adadi na kwance (compression ratios), har ma da abubuwan tsarin fayil don gano ko fayil an yi masa gyara ko an sake adanawa ta hanyar software.

Hadakar waɗannan dabarun AI yana ƙirƙirar kariya mai matakai da yawa, yana juya abin da a baya ya kasance a cikin rami mara gani zuwa majiyoyin shaidu masu ƙarfi.

Mafi Kyawun Ayyuka ga Masu Bincike

Aiwatar da AI baya nufin kawai gudu da kayan aiki. Yana nufin tsara wani yanayin aiki (workflow) wanda ke haɓaka daidaito da inganci:

  1. Samar da Asali (Baseline): ajiye cikakken kundin misalai na samfuran kamara da aka sani, alamu na software gyara hotuna na gama gari, da kundin hotuna na yankuna. Wannan ajiya na hanzarta dacewar daidaito da gano ba daidai ba.
  2. Ƙaumata da Hanyar Matakan Mataki (Layered Approach): Ba wata hanya daya ba ce da ke tabbatar da sakamako. Haɗa nazarin hayaniya, binciken ma’anoni, da gano deepfake don tabbatar da sakamako. Idan kayan aikin guda biyu ko fiye masu zaman kansu sun nuna irin wancan yanki, zaku iya samun tabbaci.
  3. Haɗa Ƙwararrun Mutane (Integrate Human Expertise): AI na haskaka abubuwan banbanci, amma shawarar ɗan Adam har yanzu na da muhimmanci. Horar da ƙungiyar ku don sake duba taswirar zafi da yankunan da aka yi nuni akansu, fassara sakamako a cikin mahalli kuma ku guji ƙarya mai yawa.
  4. Rubuta A kowane Mataki (Document Every Step): Samar da rajista fili. Rikota sigar kayan aiki, saitunan sigogi, da bayanan nazari. Wannan rubutun yana ƙarfafa sahihancin sakamakonku a cikin hali na shari’a ko bin ƙa’ida.
  5. Kasance Kan Ci gaba (Stay Current): Dukkan dabarun counter-forensic da kayan gano suna haɓaka cikin sauri. Yi horo akai-akai, halarta taruka na masana, da shiga cikin tattaunawa na ƙungiyoyi don sabunta hanyoyinku.

Kammalawa: Ƙarfafa Fa’ida tare da AI

Masu aikin anti-forensic na yau suna da ƙwarewa, amma ba lallai bane ku kasance a baya. Ta hanyar haɗa nazarin hayaniya da AI, gano deepfake, da geolocation bisa ga mahalli cikin kayan aikin bincikenku, kuna samun wani bambanci mai karya. Za ku gano rashin jituwa waɗanda ba za a gani da ido ba kuma ku juya hotunan masu zargi daga wuraren da za su zama tiren bincike zuwa shaidu masu amfani.

Shirya don haɓaka aikin tabbatar da hotonku? Gwada sabis na GeoClue da ke amfani da AI don hoton-geolocation. Ajiye ainihin wuri da hoton ya ɗauka, ko da an cire bayanan EXIF. Juya shaidu na gani zuwa lambobin ƙasa da wuri tare da ƙididdigar amincewa a cikin dakika, kuma ku kasance a gaba wajen yaudara.

Cire kuskuren hasashen wuri da tabbataccen hoton hoto. Yi rijista don GeoClue a yau kuma fara bayyana counter-forensics tare da ikon AI.