Landmark extraction is a computer vision technique that identifies and maps key points on a hand, face, or other object. For a hand, these landmarks can include fingertips, joints, and the palm; for a face, they may include eyes, nose, and mouth. The resulting point map turns a live image or video frame into structured data that software can compare, track, or classify.
In cyber security, landmark extraction is useful in gesture-based CAPTCHA, face authentication, and liveness detection because it helps systems verify that a real person is present and moving naturally. It can also support anti-spoofing checks by measuring motion over time and spotting inconsistencies in shape or pose. Attackers may try replay attacks with photos or videos, or use virtual camera tools to feed fake input, so landmark-based systems must be combined with server-side risk scoring and other signals.