Occlusion is anything that blocks or hides part of a face, body, object, or scene from a camera view. In computer vision, even small occlusions such as a hand, mask, hood, scarf, hair, glare, or another person can reduce how well a model detects, tracks, or identifies a subject. The blocked area may remove key features that the system relies on for face recognition, person re-identification, or object classification.
In cyber security, occlusion matters because many modern surveillance and access-control systems depend on vision pipelines. Attackers can use deliberate occlusion as a physical adversarial technique to weaken biometric matching or make tracking less reliable. Defenders must test for it, because real environments are messy: lighting changes, angles shift, and clothing or accessories can hide features unpredictably. A system that works in clean lab images may fail when faces or bodies are partially obscured, so robust evaluation must include occluded and partially visible inputs.