Face Matching
Face matching is the ability of a biometric system to confirm that two images, such as a selfie and the photograph on a government-issued ID, belong to the same person. It produces a similarity score, which is compared against a threshold to return a match or no-match decision.
Face matching is a component of biometric verification, and it should not be mistaken for liveness detection. The two answer different questions and run as separate stages in the same flow. Matching asks whether these two images correspond. Liveness asks whether the face belongs to a real person who is present right now.
In an onboarding flow, the user captures their document, takes a selfie, face matching compares the two, then liveness confirms the capture is of a live person. In an authentication flow, the document step drops away, but both stages remain distinct.
What Face Matching Cannot Do
Face matching provides no presentation attack detection (PAD), so it cannot distinguish a real face from a photograph, a mask, or a screen held up to the camera. It also provides no generative AI mitigation such as digital injection attack detection (DIAD).
This is the most consequential and most widely misunderstood limitation. A high-quality deepfake of a genuine account holder is designed to look like that person, so a matching algorithm will do exactly what it was built to do and report a strong match. The match is correct. The premise is false.
Face matching also differs from face recognition, which searches one face against a database of many rather than comparing two. iProov delivers face matching and face verification where the user is fully aware, and does not deliver face recognition for surveillance.
Both face matching and liveness are required for robust verification; face matching alone is not suitable for this task. iProov combines face matching with Dynamic Liveness® to confirm the individual is the right person, a real person, and verifying right now.Â