I want a liveness or “realness” check in the Face Detection API

I want the browser's Face Detection API to expose a liveness or "realness" confidence score alongside the existing face detection result, so that developers can build anti-spoofing features without relying on external machine-learning libraries.

The Shape Detection API's FaceDetector is currently in an experimental stage, and it does a good job of telling me whether a face is present in a frame. What it does not tell me is whether that face belongs to a real, live person or whether it is a photograph, a printed image, or a video replay being held up to the camera. This distinction is critical for any application that uses face detection as part of an identity or presence verification flow — such as a check-in system, an anti-cheat mechanism in an online exam, or a liveness gate before a sensitive action.

A livenessScore property on the DetectedFace result — or a separate liveness detection mode — would give developers a native, privacy-respecting signal they could use to flag potential spoofing attempts when consuming a live getUserMedia() stream. Keeping this in the platform means it can be implemented using on-device models, avoiding the need to send video frames to a third-party service and reducing both latency and privacy risk.