Skip to content

Injection Attack.

In the IDV context, an attack that injects pre-rendered synthetic media (a deepfake video, a synthetic face capture) directly into the camera or sensor pipeline at the OS, driver, or virtual-camera level — bypassing the lens entirely.

Presentation attacks (holding up a photo) are largely defeated by modern liveness detection. Injection attacks (feeding deepfake video directly into the camera API) bypass the camera entirely and bypass most liveness defenses with it. The defense moves outside the capture — device attestation, hardware-signed frames, anomaly detection on capture timing and metadata — none trivial to deploy at scale.

Common questions

What's the difference between a presentation attack and an injection attack?

Can liveness detection catch injection attacks?

What defends against camera-feed injection?

Related terms

In the guides

Last updated 2026-05-15.