1、NIST Face Morph Detection Evaluation(FATE MORPH)Mei NganNational Institute of Standards and Technology(NIST)U.S.Department of CommerceInternational Face and Fingerprint Performance Conference 2025April 3rd,20251:1 VERIFICATION1:N SEARCHTWINS DISAMBIGUATIONFACE IN VIDEO 2024MORPH DETECTIONPADAGE ESTI
2、MATIONTWO PEOPLE IN ONE FACE?SUBVERSIVE PHOTO?HOW OLD?OLD ENOUGH?SAME PERSON OR NOT?WHO?WHERE?WHEN?SAME PERSON,OR TWIN?1:N ON NON-COOP PEOPLEQUALITY+DIAGNOSTICSHOW BAD IS THIS PHOTOFRTEFACE RECOGNITIONRECOGNITION TECHNOLOGY EVALUATIONFATEFACE ANALYSIS ANALYSIS TECHNOLOGY EVALUATIONRECOGNITION:WHO IS
3、 IN AN IMAGENISTs morph detection benchmarkFATE MORPH 2018-present S-MAD D-MAD FR vulnerability Image reversalANALYSIS:ABOUT IN AN IMAGE2Face morphingSubject ASubject B3Image Source:NISTFace morphing generates an image that visually resembles both contributing subjectsMorph(A+B)Problem:face recognit
4、ion matches both personsMorph(A+B)Subject ASAME PEOPLE Both people can authenticate against a morph All modern face recognition algorithms tested by NIST and operational matchers tested by parts of the U.S.Government are vulnerable to morphsFace RecognitionMorph(A+B)Subject BSAME PEOPLEFace Recognit
5、ionThreat:one document,two usersAccomplice(passport owner)Attacker(other identity)PASSPORT APPLICATION+Morph=A+BPrinted on 2in x 2in photo paper andmailed to passport officePassport office scans printed photo,conducts human examination,Ultimately,passport gets issued with morphed photo on itImage So
6、urce:NIST5Some U.S.passport applications are susceptible to manipulation of user-submitted photos.Other countries also accept user-submitted photos for identity credential applications.Risk#1:Organizations that accept user-submitted photos for enrollment are at risk of morphing attack.Risk#2:Organiz