Skip to content

Aries RFC 0231: Biometric Service Provider

Summary

Biometric services for Identity Verification, Authentication, Recovery and other use cases referred to in Aries RFCs including DKMS.

Motivation

Biometrics play a special role in many identity use cases because of their ability to intrinsically identify a unique individual, but their use depends on a variety of factors including liveness, matching accuracy, ease of acquisition, security and privacy. Use of biometrics is already well established in most countries for domestic and international travel, banking and law enforcement. In banking, know-your-customer (KYC) and anti-money laundering (AML) laws require some form of biometric(s) when establishing accounts.

In this specification, we characterize the functions and schema that biometric service providers (BSPs) must implement to ensure a uniform interface to clients: wallets and agents. For example, current Automated Biometric Information Systems (ABIS) and other standards (IEEE 2410, FIDO) provide a subset of services but often require proprietary adaptors due to the fragmented history of the biometric market: different modalities (face, fingerprint, iris, etc.) require different functions, schema, and registration information. More recently, standards have begun to specify functions and schema across biometric modalities. This specification will adopt these approaches and treat biometric data within an encrypted envelope across modalities.

Tutorial

One goal of the Biometric Service Provider (BSP) specification is to allow for self-sovereign biometric credentials in a holder's wallet or cloud agent trusted by issuers and verifiers:

bsp_arch1

An issuer may collect biometric information from a holder in order to issue credentials (biometric or not). Likewise, a verifier may require biometric matching against the holder's credentials for authentication. In either case, issuers, holders and verifiers may need to rely on 3rd party services to perform biometric matching functions for comparison to authoritative databases.

Basics

In general, biometrics are collected during registration from a person and stored for later comparisons. The registration data is called the Initial Biometric Vector (IBV). During subsequent sessions, a biometric reading is taken called the Candidate Biometric Vector (CBV) and "matched" to the IBV:

bsp_arch2

Both the IBV and CBV must be securely stored on a mobile device or server often with the help of hardware-based encryption mechanisms such as a Trusted Execution Environment (TEE) or Hardware Security Module (HSM). The CBV is typically ephemeral and discarded (using secure erasure) following the match operation.

If the IBV and/or CBV are used on a server, any exchange must use strong encryption between client and server if transmitted over public or private networks in case of interception. Failure to properly protect the collection, transmission, storage and processing of biometric data is a serious offense in most countries and violations are subject to severe fines and/or imprisonment.

Example: Aadhaar

The Aadhaar system is an operational biometric that provides identity proofing and identity verification services for over 1 billion people in India. Aadhaar is comprised of many elements with authentication as the most common use case:

bsp_aadhaar

Authentication Service Agents (ASAs) are licensed by the Government of India to pass the verification request via secure channels to the Unique Identification Authority of India (UIDAI) data centre where IBVs are retrieved and matched to incoming CBVs from Authentication User Agencies (AUA) that broker user authentication sessions from point-of-sale (PoS) terminals:

Use Cases

A Biometric Service Provider (BSP) supports the following use cases. In each case, we distinguish whether the use case requires one-to-one (1:1) matching or one-to-many (1:N) matching:

  1. Device Unlocking - primarily introduced to solve the inconvenience of typing a password into a small mobile device, face and single-digit fingerprint was introduced to mobile devices to protect access to the device resources. This is a 1:1 match operation.

  2. Authentication - the dominant use case for biometrics. Users must prove they sufficiently match the IBV created during registration in order to access local and remote resources including doors, cars, servers, etc. This is a 1:1 match operation.

  3. Identification - an unknown person presents for purposes of determining their identity against a database of registered persons. This is a 1:N match operation because the database(s) must be searched for all IBVs of matching identities.

  4. Identity Verification - a person claims a specific identity with associated metadata (e.g., name, address, etc.) and provides a CBV for match against that person's registered biometric data to confirm the claim. This is a 1:1 match operation.

  5. Identity Proofing - a person claims a specific identity with associated metadata (e.g., name, address, etc.) and provides a CBV for match against all persons in database(s) in order to determine the efficacy of their claims and any counter-claims. This is a 1:N match operation because the database(s) must be searched for all IBVs of matching identities.

  6. Deduplication - given a CBV, match against IBVs of all registered identities to determine if already present or not in the database(s). This is a 1:N matching operation.

  7. Fraud prevention - A match operation could return confidence score(s) (0..1) rather than a simply boolean. Confidence score(s) express the probability that the candidate is not an imposter and could be used in risk analysis engines. This may be a use case for BSP clients.

  8. Recovery - Using biometric shards, a process using one's biometrics to recover lost private keys associated with a credential is possible using secret sharing. This may be a use case for BSP clients.

The previous diagram describing the IBV and CBV collection and matching during registration and presentation did not specify where the IBV is persisted nor where the match operation is performed. In general, we can divide the use cases into 4 categories depending on where the IBV is persisted and where the match must occur:

bsp_2x2

Mobile-Mobile: The IBV is stored on the mobile device and the match with the CBV occurs on the mobile device

Mobile-Server: The IBV is stored on the mobile device, but the match occurs on a server

Server-Mobile: The IBV is stored on a server, but the match occurs on a mobile device

Server-Server: The IBV is stored on a server and the match occurs on a server

Use case 1: Identity Proofing

Use case 2: Recovery

Reference

The NIST 800-63-3 publications are guidelines that establish levels of assurance (LOA) for identity proofing (Volume A), authentication (Volume B), and federation (Volume C). The Biometric Service Provider (BSP) specification deals primarily with identity proofing and authentication.

A common misconception is that a biometric is like a password, but cannot be replaced upon loss or compromise. A biometric is private but not secret, whereas a password is secret and private. Used correctly, biometrics require presentation attack detection (PAD), also called liveness, to ensure that the sensor is presented with a live face, fingerprints, etc. of a subject rather than a spoof, i.e., a photo, fake fingertips, etc. Indeed, NIST 800-63-3B requires presence of a person in front of a witness for Identity Assurance Level 3 (IAL3) in identity proofing use cases.NIST characterizes the identity proofing process as follows:

bsp_nist_idproofing

Remote use of biometrics is increasing as well to streamline on-boarding and recovery processes without having to present to an official. NIST 800-63-3A introduced remote identity proofing for IAL2 in 2017 with some form of PAD strongly recommended (by reference to NIST 800-63-3B). Typically, additional measures are combined with biometrics including knowledge-based authentication (KBA), risk scoring and document-based verification to reduce fraud.

Protection

Biometric data is highly sensitive and must be protected wherever and whenever it is collected, transmitted, stored and processed. In general, some simple rules of thumb include:

  • Protect biometric data at collection (sensor safeguards)
  • Protect biometric data at rest (special hardware, TPM/TEE, database encryption)
  • Protect biometric data in transit (encrypted communications)
  • Protect biometric data during match (volatile memory protections)
  • Never log biometric data!
  • The Candidate Biometric Vector (CBV) is ephemeral

Issues

  • Should define biometric verifiable credential schema(s)
  • Biometric modality agnostic
  • Accommodate Biometric shards
  • Integrate with Ursa crypto
  • Integrate with service endpoint model
  • Allows Issuers, Verifiers, and Holder wallets & agents to invoke services like:
  • Registration
  • Matching
  • Deduplication
  • Verification
  • Provides new services
  • Fuzzy matching
  • Shard management (for DKMS) including Horcrux-based storage options
  • Holder-specific biometric matching “machine” (using ZK-STARKs)
  • Compatible with trust relationships
  • Supports DID connections/Trust relationships (Holder ⟷ BSP ⟷ Verifier)

Drawbacks

Biometrics are explicitly required in many global regulations including NIST (USA), Aadhaar (India), INE (Mexico), and RENIEC (Peru) but also standardized by international organizations for travel (IATA) and finance (FATF).

Rationale and alternatives

By addressing biometrics, we seek to provide explicit guidance to developers who will undoubtedly encounter them in many identity credentialing and authentication processes.

Prior art

Several biometric standards exist that provide frameworks for biometric services including the FIDO family of standards and IEEE 2410. Within each biometric modality, standards exist to encode representations of biometric information. For example, fingerprints can be captured as raw images in JPEG or PNG format but also represented as vectors of minutiae encoded in the WSQ format.

Unresolved questions

  • What is the threat model?
  • Alignment with other RFCs
  • Distributed Key Management RFC (Open Issues and Future Work section)
  • Credential Fraud RFC (Threat model, Patterns & Anti-Patterns)
  • Relation to
  • IEEE 2410 (BOPS)
  • FIDO and new FIDO IDWG (Identity Verification)

Implementations

The following lists the implementations (if any) of this RFC. Please do a pull request to add your implementation. If the implementation is open source, include a link to the repo or to the implementation within the repo. Please be consistent in the "Name" field so that a mechanical processing of the RFCs can generate a list of all RFCs supported by an Aries implementation.

Name / Link Implementation Notes