Glossary · Technology
False positive
In face recognition, a false positive is a match reported between two faces that belong to different people. Its mirror image, a false negative, is a missed match between faces of the same person. The rates of both are set by the decision threshold, and reducing one raises the other, which is why testing reports them together.
Updated November 6, 2026 · 2 min read
How are the two errors related?
Every comparison produces a similarity score; a threshold decides whether it counts as a match. Lower the threshold and more genuine pairs pass, but so do more impostor pairs. Raise it and the reverse. Testing therefore reports the false match rate at a fixed false non-match rate, or plots the two against each other, rather than quoting a single accuracy number.
| Measure | Meaning | Typical use |
|---|---|---|
| FMR (false match rate) | Share of impostor comparisons wrongly accepted | 1:1 verification |
| FNMR (false non-match rate) | Share of genuine comparisons wrongly rejected | 1:1 verification |
| FPIR (false positive identification rate) | Share of searches returning a wrong candidate above threshold | 1:N identification and search |
| FNIR (false negative identification rate) | Share of searches missing the true person | 1:N identification and search |
Why are false positives common in web search?
Because the index is enormous and the threshold is low on purpose. A search engine would rather show twenty candidates, including look-alikes, and let the user judge, than show nothing. That is the doppelganger effect by design. In one-to-many matching the expected number of false positives grows with the gallery size at any fixed threshold.
Where does it go wrong in practice?
- A false positive treated as identification: wrongful accusations, wrongful arrests.
- A false negative treated as proof of absence: "no results, so my photo is not online" is never a safe conclusion.
- Uneven error rates across demographic groups, measured by NIST in 2019 and since, which means the same threshold is not equally reliable for everyone.
How should a user read results?
As candidates. Confirm with evidence that does not come from the face: the name, date and context on the source page, or a second independent engine. Independent error rates for consumer face search engines do not exist, because they do not submit to NIST testing; treat any published accuracy figure from a vendor as unverified.
Questions
What is a good false positive rate?
It depends on the use. A border gate wants FMR near one in a million; a web search accepts far more because a person reviews the list.
Can an engine have zero false positives?
Only by rejecting almost everything, which produces false negatives instead. The two trade off.
Sources
- 1.NIST FRTE 1:N identification, error metrics (checked December 1, 2025)
- 2.ISO/IEC 19795-1, Biometric performance testing and reporting (checked May 1, 2021)
Entry reviewed November 6, 2026. Found an error? Tell us.
Related entries
- Term Similarity score Similarity score defined: the number from comparing two face embeddings, what it does and does not mean, thresholds, and why a high score is a lead, not proof.
- Term Doppelganger effect The doppelganger effect: why a face search returns look-alikes, how index size and photo quality make it worse, and how to verify a match with non-face evidence.
- Term One-to-many matching (1:N) One-to-many (1:N) face matching defined: a probe face against a gallery of many, how it differs from 1:1 verification, and why errors grow with gallery size.
- Concept Accuracy, bias and testing How face recognition accuracy is measured, what NIST FRTE reports found on error rates and demographic differences, and why marketing claims cannot be checked.
- Organisation National Institute of Standards and Technology (NIST) The US standards agency whose FRTE and FATE programmes publish the only continuing independent tests of face recognition accuracy, including demographic effects.
Cite this entry
FaceSearch Wiki editors. "False positive". FaceSearch Wiki, updated November 6, 2026. https://facesearch.wiki/glossary/false-positive/