Method and apparatus for recognizing illegal behavior in unattended scenario

Methods, apparatuses, and devices, including computer programs encoded on computer storage media, for recognizing an unauthorized behavior are provided. One of the methods includes: obtaining, through machine vision, body movement pattern data of a user; obtaining feature data of an object; comparing the body movement pattern data with a preset body movement pattern; and determining that the user's behavior is unauthorized in response to the body movement pattern data matching the preset body movement pattern and that: a distance between the user's body and the object is within a distance threshold over a preset length of time, the attribute category of the object is a preset category, and the external surface image data does not match a default external surface image of the object. The feature data may include: position data of the object based on a radio frequency identification tag on the object, attribute category data of the object, and external surface image data of the object.