Face Search Glossary
Reverse face search borrows vocabulary from machine learning and from privacy law, and the two use some of the same words differently. These are the terms that come up when reading results or reading the rules, defined plainly, with a note on why each one matters in practice.
Last reviewed
Most misunderstood
- An embedding is a description of a face, not a picture of one, and cannot be turned back into a photo.
- A confidence score measures similarity between two faces, not the probability that you are right.
- A false positive is two different people scoring as a match, and it is the error that causes harm.
- A biometric identifier is a legal category, and it is what most face-privacy law regulates.
- Reverse image search and reverse face search are different operations on the same file.
What is a face embedding?
An embedding is the numeric description a model produces from a face: a fixed-length list of numbers capturing the geometric relationships between features. It is not a compressed photograph and cannot be reversed back into one. Two photos of the same person produce embeddings that sit close together in that numeric space, and two photos of different people produce embeddings that sit further apart. Everything a face search does is measuring that distance.
What is a confidence score?
A confidence score expresses how similar two faces are, on a scale the system defines. It is easy to misread as a probability that the match is correct, and it is not that. A score of 90 does not mean a 90 percent chance this is the same person. It means the two embeddings are very close by the model's measure, which is strong evidence in a good photo and much weaker evidence in a poor one. The photo quality is baked into how much a score is worth.
What is a false positive?
A false positive is when two different people are scored as a match. This is the error that does real damage, because it attaches a stranger's identity to someone. Lookalikes, siblings, and low-quality images all raise the rate. It is the reason a match should be treated as a lead to verify against other evidence rather than as a conclusion, and the reason no reverse face search engine can honestly claim to confirm identity.
What is a false negative?
A false negative is when the same person is present but not returned as a match. It is the quieter failure and the more common one in everyday use, caused by a bad angle, a filter, an obstruction, or simply a face that is not on any indexed public page. Because false negatives are common, an empty result is weak evidence that someone has no online presence.
What is a biometric identifier?
A biometric identifier is a legal category, not a technical one: data derived from a physical characteristic that can single out one individual. A face embedding usually falls inside it. This matters because most face-privacy law, including GDPR Article 9 and Illinois BIPA, regulates the processing of biometric identifiers rather than the act of looking at a photograph, so the rules can be triggered by computing the embedding at all.
What is the difference between reverse image search and reverse face search?
Reverse image search and reverse face search are two different operations on the same file. Reverse image search fingerprints the picture and finds copies of it: the same file reposted, cropped, or resized. Reverse face search discards the picture and keeps only a description of the face, then finds other photographs of that person, which may look nothing alike. Asking the wrong one is the usual reason a search seems to find nothing.
What is liveness detection?
Liveness detection decides whether a camera is seeing a real person present in front of it or a photograph, screen, or mask held up to it. It belongs to identity verification and unlocking, not to reverse face search. FaceSearch does not perform liveness detection, because it works from an image you supply rather than from a live camera. The distinction matters when reading privacy policies, since the two are regulated and discussed together but do very different things.
What is an index in this context?
The index is the searchable collection of face embeddings built from publicly reachable pages. A search compares your embedding against the index, so the index defines the boundary of what can possibly be found. Pages behind a login, blocked to crawlers, or never crawled are not in it, which is why private accounts are invisible regardless of how good your photo is.
Sources
- Face Recognition Vendor Test (FRVT) 1:1 Verification, National Institute of Standards and Technology
- Special category data, Information Commissioner's Office