Glossary

Face search glossary

Plain definitions of the terms used in face search, privacy and image law.

B

Biometric data Law: Personal data from a person's physical or behavioural traits, such as a face template, that allows or confirms their unique identification.

C2PA content credentials Technology: Cryptographically signed metadata, defined by the C2PA standard, that records the origin and edit history of an image or video so that tampering can be detected.
Catfishing Safety: Deceiving someone with a false online identity, typically built on another person's photos, for a relationship, attention or trust rather than directly for money.
Consent Law: A freely given, specific, informed and unambiguous agreement to a particular processing of personal data; explicit when the data is biometric.
Crawler Search: A program that fetches public web pages and follows links so their content, including images, can be indexed.

Data broker Privacy: A business that compiles personal information about people it has no direct relationship with, from public and commercial sources, and sells or licenses the resulting profiles.
Deep search Search: A wider or slower second pass that some engines offer beyond the default search, covering more of the index or more sources, usually for a separate credit.
Deepfake Technology: Synthetic media in which a real person's face, body or voice is replaced or generated by a machine-learning model so they appear to do or say something they did not.
DMCA takedown Law: A notice under 17 U.S.C. 512 asking a service provider to remove material that infringes the sender's copyright; applies to photos you own, not photos of you.
Doppelganger effect Technology: The appearance of unrelated look-alikes in face search results, caused by comparing one face against a very large index where some stranger always resembles the query.

EXIF metadata Privacy: Data embedded in a photo file by the camera or phone, such as date, device and GPS location, that travels with the file unless stripped.

Face detection Technology: Finding the position of faces in an image, typically as a bounding box plus landmark points, without saying whose face it is.
Face embedding Technology: A vector of numbers computed from an aligned face by a neural network, designed so that the same person's faces produce nearby vectors.
Face recognition Technology: Technology that compares faces to decide whether they belong to the same person, used for verification, identification and search.
Facial recognition search Search: Searching a large image collection, typically the public web, for a given face using face recognition; in consumer use the same thing as a reverse face search.
False positive Technology: A match reported between faces of two different people; paired with the false negative, a missed match between faces of the same person.

Impersonation Safety: An account, page or message that presents itself as a specific real person or organisation, using their name, photos or details, without their permission.
Index Search: The stored collection of face embeddings and source page addresses that a face search engine compares a query against.

Liveness detection Technology: A check that the face in front of a camera belongs to a live, physically present person rather than a photo, screen replay, mask or deepfake feed.

NCII Safety: Non-consensual intimate images: sexual or intimate photos or videos, real or synthetic, shared without the consent of the person shown.

One-to-many matching Technology: Comparing one probe face against a gallery of many enrolled faces and returning a ranked candidate list, the operation behind face search and identification systems.
Opt-out Privacy: A request to a face search engine to stop returning (and usually indexing) a person's face; it affects only that engine, not the source websites.
OSINT Search: Open-source intelligence: collecting and analysing publicly available information, including images and faces, to answer a question about a person, place or event.

Perceptual hash Technology: A compact fingerprint of an image's visual content that stays similar when the image is resized, compressed or lightly edited, used to recognise known images.
Public web Search: Pages that load for any visitor without a login; the only content a crawler can fetch and a face search engine can index.

Re-encoding Technology: Decoding an image and writing it out as a new, clean file, which removes embedded metadata and neutralises malformed input.
Reverse face search Search: A search where the query is a face: the engine finds other public photos of the same or a similar face, not copies of the same picture.
Reverse image search Search: A search where the query is an image: the engine finds copies, crops and visually similar versions of that picture on public web pages.
Romance scam Safety: Fraud in which an offender uses a false online identity and a fake relationship to obtain money, cryptocurrency or financial access from the victim.

Sextortion Safety: Blackmail in which an offender threatens to publish intimate images of a person unless the person pays, sends more images or does something else the offender demands.
Similarity score Technology: The output of comparing two face embeddings; higher means closer resemblance under the model, not a probability of identity.
Source page Search: The public web page on which a matching image was found; the part of a result that identifies the publisher and the removal route.