Biometrics Beyond Facial Recognition
How fingerprints, palmprints, iris recognition, voice and speaker recognition, DNA, gait, and multimodal biometric systems can help establish or investigate identity—and what agencies should understand about identification versus verification, candidate matches, collection quality, false matches, database searches, privacy, evidence, human review, and governance.
What this explainer does
Biometrics are measurable physical or behavioral characteristics that can be used to recognize or help verify an individual's identity. Facial recognition is only one biometric modality. Law enforcement has relied on fingerprints for generations, while modern systems increasingly support palmprints, iris images, DNA profiles, voice characteristics, and combinations of several biometric signals.
Different biometric technologies answer different questions and operate under different conditions. A fingerprint recovered from a crime scene is not the same type of input as a controlled iris capture at booking. A voice recording from an unknown caller is not equivalent to a tenprint fingerprint submission. A multimodal identification system may combine several signals rather than relying on one.
The central governance issue is therefore broader than whether biometrics are accurate. Agencies need to understand what characteristic is being measured, how it was collected, what database is searched, whether the system performs identification or verification, how uncertainty is expressed, and what independent review is required before the result affects official action.
The FBI's Next Generation Identification system currently provides criminal-justice biometric services involving fingerprints, palmprints, iris images, and facial images. The FBI describes the platform as capable of evolving multimodal functionality.
NIST's biometric work extends still further, including fingerprint, iris, face, voice, DNA, and multimodal technologies. Agencies should therefore think of biometrics as an expanding technology category rather than facial recognition as an isolated issue.
1. Overview
Biometrics turn characteristics of a person into information that can help answer an identity question.
NIST broadly describes biometrics as automated recognition based on biological or behavioral characteristics. Those characteristics can include fingerprints, palmprints, facial features, iris patterns, voice characteristics, vein patterns, and behavioral signals.
In criminal justice, biometric systems can support several different functions: booking and identity verification, comparison of latent evidence against known records, identification of unknown persons, correctional intake and release, background processes, forensic investigation, and generation of investigative leads.
The technology varies substantially by modality. Some biometric samples are intentionally captured under controlled conditions. Others are recovered from crime scenes or extracted from existing recordings. Some are relatively stable physiological traits; others depend more heavily on behavior, environment, equipment, or recording quality.
2. Identification and Verification Are Different Tasks
One of the most important biometric distinctions is between verification and identification.
The system compares a biometric sample against a specific person's enrolled reference to determine whether the sample is consistent with the claimed identity. The question is essentially: “Is this person who they claim to be?”
The system compares one biometric sample against many records in a database to identify one or more possible candidates. The question is essentially: “Who might this person be?”
Why the Difference Matters
Searching a single known reference is mathematically and operationally different from comparing one unknown sample against millions of records. A very large candidate database creates more opportunities for unrelated records to produce scores above a selected threshold.
3. Major Biometric Modalities
| Modality | Characteristic | Typical Law-Enforcement Context | Important Limitation |
|---|---|---|---|
| Fingerprints | Friction-ridge patterns of fingers | Booking, identity, latent evidence, background processes | Latent quality, distortion, partial impressions, examiner interpretation |
| Palmprints | Friction-ridge patterns of palms and related areas | Latent crime-scene evidence and biometric repositories | Quality and completeness of recovered or enrolled impressions |
| Iris | Detailed iris pattern | Controlled identification, corrections, booking, identity verification | Requires suitable capture and enrollment; not equivalent to casual eye photography |
| Voice / Speaker | Acoustic and behavioral characteristics of speech | Recorded calls, authentication, investigative comparison | Noise, channel, illness, disguise, emotion, language, and recording conditions |
| DNA | Genetic profile derived from biological material | Forensic identification, exclusion, CODIS, known-person comparison | Mixtures, contamination, transfer, interpretation, collection and laboratory issues |
| Gait | Patterns of walking or bodily movement | Video analysis and emerging investigative applications | Clothing, injury, camera angle, speed, surface, carrying objects, and limited standardization |
| Multimodal | Two or more biometric characteristics | Higher-confidence identity workflows and system interoperability | Complexity, data sharing, differing modality error rates, and combined-system governance |
4. Fingerprints
Fingerprint identification remains the best-established biometric technology in law enforcement. Automated fingerprint systems can compare known tenprint records or latent impressions against large repositories and return potential candidates for further examination.
Fingerprint technology illustrates an important principle applicable across biometrics: automated search and forensic conclusion are not necessarily the same step.
Booking or enrollment fingerprints can be collected under standardized conditions designed to capture high-quality ridge detail.
Crime-scene impressions may be partial, distorted, smudged, overlapping, or deposited under uncontrolled conditions.
Algorithms can rapidly compare ridge characteristics against large repositories and generate candidate records.
5. Palmprints and Supplemental Friction-Ridge Biometrics
Hands contain useful friction-ridge information beyond the fingertips. Palm surfaces and additional portions of fingers may leave latent impressions at crime scenes and can be enrolled into searchable biometric systems.
The FBI's Next Generation Identification system includes palmprint and supplemental-print search capabilities, allowing law enforcement to compare relevant evidence against enrolled biometric records.
Hands contacting doors, counters, windows, weapons, vehicles, or other surfaces may leave usable palm impressions.
The palm can provide additional friction-ridge information when traditional fingertip evidence is absent or incomplete.
Automated repositories can generate candidates for examination from enrolled palmprint records.
6. Iris Recognition
Iris recognition analyzes detailed patterns in the colored portion of the eye surrounding the pupil. Modern systems typically capture iris images using specialized equipment, often including near-infrared illumination, and convert those patterns into a form suitable for automated comparison.
In 2020, the FBI placed its NGI Iris Service into operation as a national iris repository and search capability for authorized criminal-justice users. The FBI has promoted iris capture particularly for controlled environments such as booking stations, correctional intake, and release processes.
Iris images can be collected without physically touching the person, potentially reducing contact during controlled identification procedures.
Properly enrolled iris images can support rapid automated searches against a biometric repository.
Iris structure can provide highly distinctive information suitable for biometric recognition.
7. Voice and Speaker Recognition
Speaker-recognition systems analyze characteristics contained in recorded speech to compare or distinguish speakers. Those characteristics can reflect both physical aspects of the vocal system and learned behavioral patterns of speech.
Voice biometrics can be useful, but recorded speech presents a particularly complex environment because the biometric signal can change with recording equipment, background noise, telephone compression, language, illness, emotion, intoxication, intentional disguise, and other conditions.
Compare a speaker against a particular enrolled or known voice reference.
Compare an unknown speaker against multiple candidate voices where the system and database support that function.
Systems may group portions of lengthy recordings by likely speaker without necessarily assigning a confirmed identity.
Synthetic Voice Complicates the Analysis
Modern generative AI can clone or synthesize voices. Investigators evaluating speaker-recognition results should therefore consider not only whether two samples sound similar, but also whether either recording may have been artificially generated or manipulated.
8. DNA as a Biometric Identifier
DNA differs substantially from image- and signal-based biometrics, but it is a biological characteristic capable of supporting identification and exclusion and is included within the broader biometric landscape.
Law enforcement has long used forensic DNA analysis to compare biological evidence with known individuals or searchable databases. Rapid DNA systems can automate development of qualifying DNA profiles from certain reference samples under defined circumstances.
Biological evidence can be compared with known reference profiles to support inclusion or exclusion.
Qualifying profiles may be searched within authorized DNA databases under governing rules and procedures.
Specialized systems can automate portions of reference-sample processing within authorized criminal-justice workflows.
9. Gait and Other Behavioral Biometrics
Not all biometrics are fixed anatomical features. Behavioral biometrics attempt to recognize people from patterns in how they act, move, speak, type, hold devices, or otherwise interact with the world.
Gait recognition analyzes patterns in how a person walks or moves. It is attractive for video analysis because it may sometimes be available when the person's face is obscured or too distant for facial recognition.
Heavy clothing, footwear, body armor, coats, or carried objects can alter visible movement.
Injury, fatigue, age, intoxication, illness, or pain can change gait.
Camera angle, frame rate, resolution, obstruction, distance, and perspective can materially affect analysis.
10. Multimodal Biometrics
A multimodal biometric system uses more than one biometric characteristic. For example, an identity workflow might combine fingerprint, face, and iris information rather than relying on a single modality.
NIST has long studied multimodal biometrics because combining modalities can improve system flexibility and potentially reduce weaknesses associated with any single signal.
Multiple biometric characteristics can provide additional information supporting an identity assessment.
Another modality may remain usable when a fingerprint sensor, face image, or other input is poor.
Properly designed fusion of multiple independent signals may improve overall recognition performance.
More Data Also Means More Governance
Combining biometric systems can increase the amount of highly identifying information retained about an individual and can expand interoperability, sharing, database-linkage, cybersecurity, and secondary-use questions.
11. A Typical Biometric Identification Workflow
12. Accuracy Is More Complicated Than a Single Percentage
Claims that a biometric system is “99 percent accurate” can be misleading without understanding what was measured, under what conditions, against what population, and at what decision threshold.
The system incorrectly treats biometric samples from different people as sufficiently similar.
The system fails to recognize two samples that actually come from the same person.
The system cannot obtain biometric information of sufficient quality to perform the intended operation.
Thresholds Matter
Many automated biometric systems produce similarity or comparison scores. The system or user then applies a threshold to determine which results receive further consideration. Raising or lowering that threshold can affect both false-match and false-non-match performance.
Database Size Matters
A one-to-many search against a very large repository is not equivalent to verifying a claimed identity against one enrolled reference. Agencies should evaluate performance under conditions resembling the intended operational use.
13. Input Quality Can Control the Result
Even a high-performing algorithm cannot recover biometric information that was never captured adequately.
A latent print, eye image, recording, or video may contain only a limited portion of the useful biometric signal.
Pressure, motion, compression, perspective, sensor characteristics, or environmental conditions may alter the observed feature.
Background sound, image artifacts, poor lighting, contamination, or unrelated data can interfere with analysis.
Sensor design, camera resolution, microphone quality, software, and capture configuration affect usable information.
Poor enrollment procedure, incorrect positioning, documentation errors, or inadequate training can reduce quality.
Compression, re-recording, editing, screenshots, exports, or platform processing can alter a biometric source.
14. Biometric Candidate Results and Investigative Leads
Some biometric technologies are designed to return candidates rather than final identification conclusions. The distinction is particularly important where automated systems search large databases.
A candidate result may justify additional investigation. It does not necessarily establish probable cause, prove identity, or justify telling a witness whom investigators believe committed an offense.
Does age, geography, physical description, or other known information fit the investigation?
Look for video, witnesses, records, digital evidence, physical evidence, location information, or other corroboration.
Investigators should actively consider information inconsistent with the candidate's involvement.
15. Privacy, Collection, Retention, and Sharing
Biometric information is unusually identifying. A password can be changed. A person's fingerprints, iris patterns, and many other biometric characteristics cannot simply be replaced after unauthorized disclosure.
Governance should therefore address not only whether biometric technology works, but whether the agency should collect the information, how long it should retain it, who may search it, and whether it may be shared or reused for other purposes.
Identify the legal authority and purpose supporting biometric acquisition.
Determine how long enrolled samples, templates, candidate results, and related records remain available.
Limit searches to authorized users and approved criminal-justice purposes.
Understand what other agencies, repositories, vendors, or systems receive biometric information.
Determine whether biometrics collected for one purpose may later be used for another.
Apply access controls, encryption, audit logging, breach response, and other safeguards appropriate to sensitive identity data.
16. Evidence, Documentation, and Discovery
When a biometric result materially affects an investigation, the agency should be prepared to reconstruct how that result was produced and how investigators used it.
| Element | What May Be Important to Preserve or Document |
|---|---|
| Source sample | The fingerprint, palmprint, iris image, recording, DNA sample, video, or other underlying biometric source. |
| Collection method | How, when, where, and by whom the biometric information was obtained. |
| Quality information | Relevant quality scores, limitations, image conditions, recording conditions, or laboratory information. |
| System used | Repository, software, algorithm, device, service, or laboratory involved in the comparison. |
| Search type | Whether the system performed verification, identification, or candidate search. |
| Search parameters | Relevant database, thresholds, filters, settings, or other configuration where material and available. |
| Candidate results | Returned records, scores, rankings, or investigative candidates where retention is appropriate. |
| Human review | Any examiner, analyst, investigator, or other review following the automated result. |
| Corroboration | Independent evidence obtained before investigators relied on a biometric candidate. |
| Exculpatory information | Evidence that contradicted, weakened, or excluded a biometric candidate. |
| Audit history | Search logs, user activity, database access, exports, or system records when relevant. |
| Expert documentation | Reports, notes, laboratory material, methodology, or validation information where expert interpretation was used. |
17. Governance Framework
Identify exactly what biometric characteristic the system collects or analyzes.
Distinguish verification, identification, forensics, access control, and investigative candidate generation.
Establish who may collect biometric information and under what authority.
Define minimum capture or sample requirements before a search is relied upon.
Identify which repositories may be searched and for what purposes.
Define how investigative candidates must be corroborated before official action.
Specify when examiner, analyst, supervisor, or other qualified review is required.
Establish rules for samples, templates, search records, candidate results, and associated case information.
Log searches, users, database access, exports, administrative activity, and other material events.
Evaluate performance under operational conditions resembling the agency's actual use.
Address biometric-data ownership, model improvement, retention, subprocessors, security, and system changes.
Monitor statutes, constitutional decisions, biometric privacy law, evidence requirements, and changing technology.
18. Questions Every Agency Should Answer
19. Where Biometrics Is Going
Identity systems may increasingly combine face, fingerprint, iris, voice, and other characteristics.
Smaller sensors may expand field acquisition of fingerprints, iris images, and other biometrics.
Systems may increasingly attempt to derive biometric information from ordinary video, audio, sensors, or behavior rather than controlled enrollment.
Facial appearance, gait, body characteristics, clothing, and other signals may increasingly be evaluated together.
Machine-learning systems may combine multiple weak indicators into broader identity assessments.
Federal, state, local, and other authorized systems may increasingly exchange standardized biometric information.
20. Key Terms
21. Related ShieldPST.ai Resources
Examine candidate generation, probe images, galleries, similarity scores, corroboration, policy, and investigative safeguards.
Open explainer →Understand voice cloning, synthetic identity, authentication, provenance, and manipulated digital evidence.
Open explainer →Explore evidence integrity, metadata, audit trails, access, retention, discovery, and secure information management.
Open explainer →Understand AI-assisted analysis, verification, hallucinations, evidence, privacy, and governance.
Open explainer →Examine object detection, tracking, search, and AI-assisted analysis of recorded video.
Open explainer →Review AI-assisted detection and protection of faces, plates, audio, documents, and other sensitive information.
Open explainer →Apply structured technology governance to biometric and other AI-supported systems.
Open resource →Research Fourth Amendment, identification, privacy, and technology decisions.
Browse case library →Return to the Shield Technology Reference Library.
Browse explainers →22. Selected Authoritative Sources
NIST's biometric research program addresses fingerprint, facial, iris, voice, DNA, multimodal systems, biometric standards, measurement, accuracy, and interoperability.
Review NIST Biometrics
FBI overview of NGI biometric services, including fingerprint, palmprint, iris, facial-image, and multimodal criminal-justice capabilities.
Review FBI NGI
FBI materials describing the national iris repository and contactless biometric identification capability available to authorized law-enforcement and criminal-justice users.
Review FBI Iris Service
2026 revision of the national biometric information-exchange standard supporting standardized interchange of fingerprint, facial, iris, and other biometric information.
Review 2026 standard
NIST research addressing systems that combine multiple biometric and recognition modalities and methods for evaluating their performance.
Review NIST Multimodal Biometrics
FBI technical resources concerning NGI iris, palmprint, rapid DNA, fingerprint, and other biometric capabilities.
Review FBI biometric modalities
23. Key Takeaways
- Biometrics extend well beyond facial recognition and include fingerprints, palmprints, iris patterns, voice characteristics, DNA, gait, and combinations of multiple modalities.
- The FBI's current Next Generation Identification system supports criminal-justice biometric searches involving fingerprints, palmprints, iris images, and facial images.
- Verification and identification are different biometric tasks: one-to-one comparison asks whether a person matches a particular reference, while one-to-many identification searches for candidates within a database.
- “Biometric match” should not be treated as a standardized conclusion. The significance of the result depends on the technology, modality, sample, database, threshold, and review process.
- Poor sample quality can materially affect even a high-performing biometric algorithm.
- False matches, false non-matches, failure-to-acquire rates, and database size should be considered when evaluating system performance.
- Candidate-generation systems should be treated as investigative tools requiring appropriate corroboration rather than automatic proof of identity.
- Voice, gait, and other behavioral biometrics can be especially sensitive to recording conditions, environment, physical condition, and intentional manipulation.
- Synthetic voices and other AI-generated media add another layer of complexity to biometric analysis and authentication.
- Multimodal biometric systems can provide additional identity information, but they also increase collection, linkage, privacy, sharing, and security concerns.
- Agencies should preserve enough information to reconstruct the source sample, search process, candidate results, human review, and corroborating investigation when biometric evidence materially affects a case.
- The governing principle is: understand exactly what biometric technology measured, what comparison it performed, what its result means, and what independent evidence supports the agency's ultimate identity conclusion.