To provide a method for performing living body detection with high accuracy without performing false recognition.

This detection device has: an image input means inputting an image of a photographic object; a variation extraction means extracting a variation of the photographic object from a plurality of images of the photographic object; a variation calculation means calculating magnitude of the variation from a candidate area extracted by the variation extraction means; and a decision means deciding that the photographic object is a living body when the magnitude of the variation measured by the variation calculation means is a predetermined threshold value or less. Preferably, the candidate area is an eye area and the periphery of the eye area.