DBIS

With the development of artificial intelligence, face recognition has become one of the most mature applications in our daily life. In real life. Deep learning-based face recognition has been widely used in applications such as: real-name person ID based on person-witness pairs, financial level face payment, and face authentication, gate verification check-in, etc.

However, at the same time, it is possible to make small pixel-level changes to the original face image through techniques such as adversarial sample generation and deep fake face generation, which can trick the face recognition system into giving incorrect results.

The goal of this thesis is to develop a novel adversarial attack method against face recognition systems and investigate possible defense mechanisms.

If you are interested in this thesis, do not hesitate to contact us via mou@dbis.rwth-aachen.de