Data Breaches
Data breaches are also common harmful outcomes of adversarial attacks. Beyond model extraction, which are attacks that are meant to gain information about the inner workings of a model like the training data, attackers use techniques to evade facial recognition models to gain unauthorized access. For example, at the Black Hat USA 2019 conference, a group of researchers from Tencent demonstrated how they were able to bypass a victim’s Apple FaceID and log into their iPhone simply by putting a pair of modified glasses on their face. Similarly, another attack on Apple’s FaceID was conducted by a Vietnamese company called Bkav. Bkav scanned a user’s face and then 3D printed a face mask. Bkav was able to trick the iPhone into thinking the actual user was trying to unlock the phone, and the phone unlocked and allowed the FaceID verification process to be evaded
