Although deepfake images and videos pose a significant threat to biometric systems used for remote identity verification, liveness technologies can help detect them. At a recent presentation hosted by the European Association for Biometrics (EAB), Ann-Kathrin Freiberg of BioID touched on basic tips for spotting deepfake videos, such as observing the transition between different areas of the face, frequency or lack of blinking, and symmetry in things like eye color or earrings. Throughout the presentation, Freiberg provided a breakdown of how deepfakes are made, their history, and real-world implications. Facial recognition technologies cannot always recognize deepfakes, but techniques used by presentation attack detection (PAD) systems can be useful. They cannot, however, help with live manipulated deepfakes, which remain highly difficult to detect.
