AI detects anomalies

Preventing manipulation in automated face recognition

Research News /

From unlocking smartphones to speeding up airport security checks: the use of automated face recognition for personal identification continues to grow. But this authentication method is vulnerable to morphing attacks: criminals can misuse it by melding two different facial images into one. A single passport featuring a photograph manipulated this way can then be used by two different people. Together with their partners, Fraunhofer research teams are developing a system that foils this type of attack using machine learning methods.