"Researchers Defeat Facial Recognition Systems With Universal Face Mask"

A team of researchers at Ben-Gurion University of the Negev and Tel Aviv University have proven that it is possible for attackers to create a face mask capable of defeating modern facial recognition systems. They validated the effectiveness of their adversarial mask in real-world experiments (CCTV use case) by printing the adversarial pattern on a fabric face mask. The facial recognition system could only identify 3.34 percent of the participants wearing the mask in these experiments, compared to a minimum of 83.34 percent with other evaluated masks. They used a gradient-based optimization process to generate a universal perturbation and mask that falsely classified each wearer as an unknown identity, even when confronted with different facial recognition models. The mask functions properly when printed on paper or fabric. While their mask is effective, it is not the only version possible. A universal perturbation's main goal is to fit any person wearing it, implying that there is a single pattern. However, the perturbation is dependent on the facial recognition model used to attack, which means that different patterns will be crafted based on the different victim models. This article continues to discuss the face mask created by the researchers that can work against many facial recognition models, as well as possible countermeasures. 

Help Net Security reports "Researchers Defeat Facial Recognition Systems With Universal Face Mask"