biometrics
Enable an agent to recognise, evaluate and govern a biometric recognition practice by recording how bodily or behavioural characteristics become identity-related evidence and what actions that evidence supports.
Research draft, second pass
A second pass drafted this model: the structure a model of this thing needs, and what is known about it in the world. The line under this one says how the second half was obtained - researched against sources, or recalled without web access, in which case nothing here was read anywhere and every claim is a lead to verify. Unreviewed either way.
recalled by Codex without web access - no source was read
Researched by: Codex
Purpose and description
Enable an agent to recognise, evaluate and govern a biometric recognition practice by recording how bodily or behavioural characteristics become identity-related evidence and what actions that evidence supports.
Biometrics is the automated recognition of individuals based on biological or behavioural characteristics, typically by comparing captured samples with previously enrolled reference data.
It can be Classify a proposed use as verification, identification or duplicate-enrolment detection and test whether its identity claims exceed the evidence.; Assess whether a sample and reference are suitable for comparison under the documented acquisition and compatibility conditions.; Evaluate a proposed operating threshold against task-specific error evidence and consequences.; Route failed capture, ambiguous matches and suspected attacks to defined retry, review or alternative procedures.; Trace and manage biometric references through enrolment, updates, compromise response and deletion.; Check whether a proposed search, linkage or downstream action falls within the recorded permitted use..
Distinguishing features
The practice uses a biological or behavioural characteristic to recognise an individual; merely measuring pulse, height or movement does not establish that it is biometrics in this model's sense.
Verification compares a probe against a claimed identity's reference, while identification searches a reference population; the declared comparison task must distinguish these cases.
A biometric comparison supplies evidence of correspondence between observations; establishing a person's civil identity additionally requires an evidenced identity binding.
Possession of a card or knowledge of a password is not itself biometric, even when the same transaction also includes a biometric comparison.
Detection of a face, voice or fingerprint without individual recognition belongs to an adjacent sensing task, although it may be a component of a biometric workflow.
Scope
+ The recognition purpose, target population and distinction between verification, identification and duplicate-enrolment detection
+ Biometric modalities, acquisition conditions and transformation of samples into reference representations
+ Comparison scores, thresholds, candidate selection and handling of uncertain or unsuccessful outcomes
+ Performance evidence, presentation attacks and other threats to biometric evidence
+ Reference lifecycle, linkability, affected-person recourse and boundaries on use
- Biometry as statistical analysis of biological populations without individual recognition
- Clinical measurements and diagnoses that do not serve a recognition purpose
- General identity proofing and credential issuance beyond their interfaces with biometric enrolment
- Complete sensor hardware specifications and manufacturing processes
- General surveillance operations, policing policies and access-control systems beyond their use of biometric evidence
Characteristics
- Recognition task
- verification; identification; duplicate-enrolment detection; other explicitly defined task Determines what is compared and what a successful result can establish.
- Biometric modality
- fingerprint; face; iris; voice; gait; other specified modality; multimodal combination Connects recognition to its acquisition requirements, variability and potential attack methods.
- Identity binding
- reference linked to a verified identity, claimed identity, pseudonym or unknown individual, with supporting provenance Separates recognition of the same individual from assurance about who that individual is.
- Acquisition context
- cooperative or noncooperative; attended or unattended; contact or contactless; sensor and environmental conditions Defines the circumstances under which capture and evaluation evidence apply.
- Reference representation
- retained sample; extracted template; protected representation; combination, with format and version Affects compatibility, exposure, migration and possibilities for replacement after compromise.
- Comparison score and operating threshold
- implementation-defined score scale and direction; threshold values tied to task and version Makes the rule translating a comparison into an outcome inspectable without treating a score as an identity probability.
- Evaluation error profile
- task-appropriate error rates with denominators, uncertainty, population, conditions and operating point Allows assessment of erroneous matches, missed matches and the limits of reported performance.
- Capture and enrolment failure
- failure-to-acquire and failure-to-enrol rates, each with a stated denominator and retry policy Records people and attempts excluded before comparison metrics are calculated.
- Reference lifecycle status
- pending; active; superseded; suspended; compromised; deletion pending; deleted with verification Controls whether a reference remains usable and whether lifecycle actions have actually completed.
- Permitted decision
- recognition outcome linked to an allowed action, governing authority, review requirement and fallback Prevents a matching result from silently becoming unrestricted authority to act.
Also called
Where this came from
wikidata · CC0 1.0
Also registered as vr.tr.biometrics
Drafted structure
Bundle to layer to finding to question, as the second pass will find it: 6 bundles · 11 layers · 18 findings · 30 questions.
Recognition meaning and identity Establishes which sense of biometrics is intended and what identity-related proposition a deployment attempts to support.
A model must distinguish biological measurement, individual recognition and proof of identity before interpreting any match.
Recognition task
Defines the comparison problem and the population against which it operates.
Task and search boundary
Record whether the operation checks a claimed identity, searches for an individual or detects repeated enrolment, including whether an individual may be absent from the reference population.
- Does this use verify a claimed identity, identify a person in a gallery or detect duplicate enrolments? definition
- Can the observed individual be absent from the gallery, and how is that possibility represented in the result? boundary
Identity binding
Separates biometric correspondence from the identity assertion attached to a reference.
Reference identity provenance
Record how the enrolled reference became associated with a person, account or pseudonym and which assertions that association supports.
- What evidence links the enrolled biometric reference to its identity label, and who established that link? provenance
- Does a match support continuity with an enrolled individual, a named identity claim or another explicitly limited assertion? boundary
Biometric acquisition and representation Describes the characteristic observed, the capture process and the representation used for recognition.
A modality name alone cannot establish that an observation is usable or comparable with an enrolled reference.
Modality and capture
Records the physical or behavioural signal and conditions that shape its acquisition.
Capture suitability
Record modality-specific acquisition conditions, quality criteria and barriers to obtaining a usable sample.
- Which characteristic is captured, using which sensor and protocol, and with what required participation from the person? definition
- How is sample quality assessed, and which environmental, bodily or behavioural conditions cause rejection or require an alternative? measurement
Samples and templates
Traces captured observations into stored references and comparison inputs.
Representation lineage
Record retained samples, feature extraction, template protection and version dependencies, including the lineage of each reference update.
- Which raw samples and derived representations are retained, and which extraction and protection versions produced them? provenance
- What evidence establishes compatibility between the enrolled reference and the current probe-processing pipeline? boundary
Comparison and performance Makes biometric decision rules and the evidence for their performance explicit.
An agent needs task-specific error evidence and an operating point to judge what a recognition result warrants.
Scores and outcomes
Connects comparisons to thresholds, candidate lists and unresolved results.
Operating rule
Record score semantics, thresholds and any fusion or ranking rules that determine the reported outcome.
- What does the comparison score mean, which direction indicates stronger correspondence, and what threshold produces each outcome? measurement
- How are multiple modalities, repeated attempts, competing candidates and inconclusive comparisons combined or resolved? action
Evaluation and transfer
Assesses whether measured performance applies to the intended deployment.
Task-specific error evidence
Record evaluation provenance, metric definitions and uncertainty across relevant populations, capture conditions and gallery sizes.
- What are the task-appropriate false-match, missed-match and acquisition or enrolment failure measures, with their denominators and uncertainty? measurement
- How do the evaluation population, gallery size, sensor conditions and system version differ from the proposed deployment? boundary
- Which population groups and operating conditions were evaluated separately, and where is evidence insufficient to assess performance differences? provenance
Biometric evidence integrity Examines whether samples, references and results can be trusted under the relevant attack conditions.
A strong similarity score does not by itself establish an authentic capture, an intact pipeline or participation by the intended person.
Presentation and capture attacks
Records attacks at acquisition and the evidence supporting capture safeguards.
Presentation attack coverage
Record tested attack presentations and the limits of presentation attack detection, distinguishing these from broader claims about identity or willingness.
- Which modality-specific attack presentations were tested, against which sensor and detector versions? provenance
- What attack acceptance and bona fide rejection measures were observed, and which attack types remain untested? measurement
Pipeline and reference attacks
Addresses substitution or manipulation beyond the sensor surface.
End-to-end evidence integrity
Record safeguards and evidence for enrolment integrity, sample transport, reference access and delivery of comparison results.
- Where could a sample, reference or match result be injected, replayed or substituted, and what verified controls cover those points? boundary
- What response is available when a reference is exposed or manipulated, and what evidence shows that a replacement limits reuse of the compromised representation? action
Use, lifecycle and recourse Connects biometric processing to permitted purposes, reference management and consequences for affected people.
Recognition evidence becomes actionable only within a defined use, with workable handling of exclusion, error and retained biometric data.
Purpose, retention and linkage
Records which biometric uses and data relationships are authorised throughout the reference lifecycle.
Permitted biometric processing
Record the authority for collection and comparison, permitted galleries and linkages, and enforceable retention and deletion conditions.
- What recorded authority permits this collection, comparison and gallery search for the stated purpose? provenance
- May samples or templates be reused or linked across services, and what technical and organisational limits enforce that boundary? boundary
- When must samples, references and copies be updated or deleted, and how is completion verified? action
Decision consequences and recourse
Defines what follows a biometric outcome and how people can recover from failure or contest an error.
Bounded actions and alternatives
Record which outcomes permit an automated action, require additional evidence or trigger an accessible alternative and review.
- Which actions may follow a match, non-match or candidate list, and which require independent evidence or human review? action
- What usable alternative and correction process exists for a person who cannot enrol, repeatedly fails capture or contests a biometric association? action
Evidence and external alignment What the world already says about this thing, gathered so the model can be checked against it.
A model that cannot be lined up against existing standards, identifiers and practice cannot be adopted by anyone who already uses them.
Reported evidence
Findings from the breadth pass, kept separate from the structural claims.
Check these first
Recalled without web access and unsourced; every item is a lead to verify.
- The registry supplies no sense; this answer selects automated personal recognition rather than the historical biological-statistics sense.
- Standards are named from recall; editions, current status and applicability require checking.
- Performance claims require task-specific testing; comparison error rates cannot be directly substituted for end-to-end authentication or identification error rates.
- Which of these check these first hold for the sense of biometrics this model covers, and on what evidence? provenance
Kinds and varieties
Recalled without web access and unsourced; every item is a lead to verify.
- Fingerprint recognition
- Face recognition
- Iris recognition
- Voice recognition
- Behavioural recognition using gait, typing or interaction patterns
- Multimodal recognition combining multiple biometric characteristics
- Which of these kinds and varieties hold for the sense of biometrics this model covers, and on what evidence? provenance
Standards and regulation
Recalled without web access and unsourced; every item is a lead to verify.
- ISO/IEC 2382-37, issued by ISO and IEC, provides biometric vocabulary.
- ISO/IEC 19795 series, issued by ISO and IEC, addresses biometric performance testing and reporting.
- ISO/IEC 30107 series, issued by ISO and IEC, addresses biometric presentation attack detection.
- ISO/IEC 24745, issued by ISO and IEC, addresses biometric information protection.
- ISO/IEC 39794 series, issued by ISO and IEC, specifies extensible biometric data interchange formats.
- Which of these standards and regulation hold for the sense of biometrics this model covers, and on what evidence? provenance
Real-world use
Recalled without web access and unsourced; every item is a lead to verify.
- Verifying an enrolled user when unlocking a device or accessing a service.
- Comparing travellers with passport or border-control reference records.
- Searching fingerprint or face collections to generate investigative candidates.
- Detecting duplicate enrolments in identity systems.
- Controlling access to workplaces and restricted facilities.
- Which of these real-world use hold for the sense of biometrics this model covers, and on what evidence? provenance
Typical measurements
Recalled without web access and unsourced; every item is a lead to verify.
- False match rate - No universal typical range; depends on modality, threshold, population and test conditions. - Proportion of non-mated comparisons incorrectly declared matches
- False non-match rate - No universal typical range; depends on modality, threshold, sample quality and test conditions. - Proportion of mated comparisons incorrectly declared non-matches
- Failure-to-enrol rate - Application-dependent; requires an explicit enrolment protocol and population. - Proportion of enrolment attempts that fail
- Recognition latency - System-dependent; identification searches also depend on gallery size. - Milliseconds or seconds per transaction
- Which of these typical measurements hold for the sense of biometrics this model covers, and on what evidence? provenance
Failure modes and hazards
Recalled without web access and unsourced; every item is a lead to verify.
- False matches can associate a person with another identity; false non-matches can deny legitimate access.
- Presentation attacks, such as artificial fingerprints or replayed speech, and digital sample injection can defeat inadequately protected systems.
- Performance can vary with demographics, disability, ageing, sensor quality and capture conditions.
- Compromised biometric data can enable tracking or impersonation, while the underlying bodily characteristic is difficult to replace.
- A biometric match can be mistaken for proof of legal identity, consent or culpability, although it establishes none of these by itself.
- Which of these failure modes and hazards hold for the sense of biometrics this model covers, and on what evidence? provenance
Regional variation
Recalled without web access and unsourced; every item is a lead to verify.
- Legal treatment varies across jurisdictions and uses, particularly for policing, workplace monitoring, consent and retention.
- National identity and border systems differ in the modalities they collect and the extent to which reference data are centralised.
- Which of these regional variation hold for the sense of biometrics this model covers, and on what evidence? provenance
Neighbouring kinds and how to tell them apart
Recalled without web access and unsourced; every item is a lead to verify.
- Biometry - Biometry concerns quantitative analysis of biological phenomena; this model covers recognition of individuals, not biological statistics generally.
- Biometric verification - Verification tests a claimed identity against its reference data, usually one-to-one; biometrics also includes identification.
- Biometric identification - Identification searches across enrolled identities, usually one-to-many; it is one biometric task rather than the whole field.
- Authentication - Authentication can use passwords, cryptographic keys or possession factors; biometrics specifically uses biological or behavioural characteristics.
- Physiological monitoring - Monitoring measures bodily state; a measurement becomes biometric recognition when used to distinguish or recognise an individual.
- Which of these neighbouring kinds and how to tell them apart hold for the sense of biometrics this model covers, and on what evidence? provenance
What the second pass must settle
- Does the registry intend biometrics exclusively as automated individual recognition, or does it also include biometry and non-identifying biological measurement?
- Which standards and disciplinary definitions should anchor the boundary between biometric recognition, behavioural profiling and continuous authentication?
- Which modality-specific evidence is needed to assess stability over time, distinctiveness and accessibility without assuming that every person can enrol?
- What independent evaluations establish the deployment limits of accuracy, template protection and attack resistance across populations and operating conditions?
- Which contextual authorities, consequences and available alternatives determine acceptable error tradeoffs, retention and opportunities for recourse?