Web Authentication: An API for accessing Scoped Credentials

1. Introduction

This section is not normative.

This specification defines an API enabling the creation and use of strong, attested, cryptographic scoped credentials by web applications, for the purpose of strongly authenticating users. A scoped credential is created and stored by an authenticator at the behest of a Relying Party, subject to user consent. Subsequently, the scoped credential can only be accessed by web origins belonging to that Relying Party. This scoping is enforced jointly by conforming User Agents and authenticators. Additionally, privacy across Relying Parties is maintained; Relying Parties are not able to detect any properties, or even the existence, of credentials scoped to other Relying Parties.

Relying Parties employ the Web Authentication API during two distinct, but related, ceremonies involving a user. The first is Registration, where a scoped credential is created on an authenticator, and associated by a Relying Party with the present user’s account (the account may already exist or may be created at this time). The second is Authentication, where the Relying Party is presented with a WebAuthn Assertion proving the presence and consent of the user who registered the scoped credential. Functionally, the Web Authentication API comprises two methods (along with associated data structures): makeCredential() and getAssertion(). The former is used during Registration and the latter during Authentication.

Broadly, compliant authenticators protect scoped credentials, and interact with user agents to implement the Web Authentication API. Some authenticators may run on the same computing device (e.g., smart phone, tablet, desktop PC) as the user agent is running on. For instance, such an authenticator might consist of a Trusted Execution Environment (TEE) applet, a Trusted Platform Module (TPM), or a Secure Element (SE) integrated into the computing device in conjunction with some means for user verification, along with appropriate platform software to mediate access to these components' functionality. Other authenticators may operate autonomously from the computing device running the user agent, and be accessed over a transport such as Universal Serial Bus (USB), Bluetooth Low Energy (BLE) or Near Field Communications (NFC).

1.1. Use Cases

The below use case scenarios illustrate use of two very different types of authenticators, as well as outline further scenarios. Additional scenarios, including sample code, are given later in §10 Sample scenarios.

1.1.1. Registration

• On a phone:

  • User navigates to example.com in a browser and signs in to an existing account using whatever method they have been using (possibly a legacy method such as a password), or creates a new account.

  • The phone prompts, "Do you want to register this device with example.com?"

  • User agrees.

  • The phone prompts the user for a previously configured authorization gesture (PIN, biometric, etc.); the user provides this.

  • Website shows message, "Registration complete."

1.1.2. Authentication

• On a laptop:

  • User navigates to example.com in a browser, sees an option to "Sign in with your phone."

  • User chooses this option and gets a message from the browser, "Please complete this action on your phone."

• Next, on their phone:

  • User sees a discreet prompt or notification, "Sign in to example.com."

  • User selects this prompt / notification.

  • User is shown a list of their example.com identities, e.g., "Sign in as Alice / Sign in as Bob."

  • User picks an identity, is prompted for an authorization gesture (PIN, biometric, etc.) and provides this.

• Now, back on the laptop:

  • Web page shows that the selected user is signed-in, and navigates to the signed-in page.

1.1.3. Other use cases and configurations

A variety of additional use cases and configurations are also possible, including (but not limited to):

• User navigates to example.com on their laptop, is guided through a flow to create and register a credential on their phone.

• A Relying Party prompts the user for their authorization gesture in order to authorize a single transaction, such as a payment or other financial transaction.

2. Conformance

This specification defines criteria for a Conforming User Agent: a User Agent MUST behave as described in this specification in order to be considered conformant. Conforming User Agents MAY implement algorithms given in this specification in any way desired, so long as the end result is indistinguishable from the result that would be obtained by the specification’s algorithms. A conforming User Agent MUST also be a conforming implementation of the IDL fragments of this specification, as described in the “Web IDL” specification. [WebIDL-1]

This specification also defines a model of a conformant authenticator (see §5 WebAuthn Authenticator model). This is a set of functional and security requirements for an authenticator to be usable by a Conforming User Agent. As described in §1.1 Use Cases, an authenticator may be implemented in the operating system underlying the User Agent, or in external hardware, or a combination of both.

2.1. Dependencies

This specification relies on several other underlying specifications.

HTML5

The concept of origin and the Navigator/dfn> interface are defined in [HTML5].

Web IDL

Many of the interface definitions and all of the IDL in this specification depend on [WebIDL-1]. This updated version of the Web IDL standard adds support for Promises, which are now the preferred mechanism for asynchronous interaction in all new web APIs.

DOM

DOMException and the DOMException values used in this specification are defined in [DOM4].

Web Cryptography API

The AlgorithmIdentifier type and the method for normalizing an algorithm are defined in Web Cryptography API §algorithm-dictionary.

The CryptoKey type for representing cryptographic keys is defined in Web Cryptography API §cryptokey-interface.

The JsonWebKey dictionary for representing cryptographic keys is defined in Web Cryptography API §JsonWebKey-dictionary.

Base64url encoding

The term Base64url Encoding refers to the base64 encoding using the URL- and filename-safe character set defined in Section 5 of [RFC4648], with all trailing '=' characters omitted (as permitted by Section 3.2) and without the inclusion of any line breaks, whitespace, or other additional characters. This is the same encoding as used by JSON Web Signature (JWS) [RFC7515].

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].

3. Terminology

Assertion

See WebAuthn Assertion.

Attestation Certificate

A X.509 Certificate for a key pair used by an Authenticator to attest to its manufacture and capabilities. The Authenticator uses the attestation private key to sign the Relying Party-specific public key (and additional data) it generates and returns upon invocation via the authenticatorMakeCredential operation.

Authentication

The ceremony where a user, and the user’s computing device(s) (containing at least one authenticator) work in concert to cryptographically prove to an Relying Party that the user controls the private key associated with a previously-registered scoped credential (see Registration). Note that this includes employing user verification.

Authenticator

A cryptographic device used by a WebAuthn Client to (i) generate a scoped credential and register it with a Relying Party, and (ii) subsequently used to cryptographically sign and return, in the form of an WebAuthn Assertion, a challenge and other data presented by a Relying Party (in concert with the WebAuthn Client) in order to effect authentication.

Authorization Gesture

Essentially the same as user verification.

Ceremony

The concept of a ceremony [Ceremony] is an extension of the concept of a network protocol, with human nodes alongside computer nodes and with communication links that include UI, human-to-human communication and transfers of physical objects that carry data. What is out-of-band to a protocol is in-band to a ceremony. In this specification, Registration, Authentication, and user verification are ceremonies.

Client

See Conforming User Agent.

Conforming User Agent

A user agent implementing, in conjunction with the underlying platform, the Web Authentication API and algorithms given in this specification, and handling communication between Authenticators and Relying Parties.

eTLD+1

Also known as a Registered Domain [PSL], an eTLD+1 is an effective Top-Level Domain Name (eTLD), plus the next domain name label, proceding from right to left. An eTLD is also known as a public suffix [RFC7719].

Registration

The ceremony where a user, a Relying Party, and the user’s computing device(s) (containing at least one authenticator) work in concert to create a scoped credential and associate it with the user’s Relying Party account. Note that this includes employing user verification.

Relying Party

The entity whose web application utilizes the Web Authentication API to register and authenticate users. See Registration and Authentication, respectively.

Note: While the term Relying Party is used in other contexts (e.g., X.509 and OAuth), an entity acting as an Relying Party in one context is not necessarily an Relying Party in others.

Relying Party Identifier

RP ID

A Relying Party Identifier is derived from a Relying Party’s web origin’s hostname by computing the hostname’s eTLD+1.

Scoped Credential

Generically, a credential is data one entity presents to another in order to authenticate the former’s identity [RFC4949]. A WebAuthn scoped credential is a { identifier, type } pair identifying authentication information established by the authenticator and the Relying Party, together, at registration time. The authentication information consists of an asymmetric key pair, where the public key portion is returned to the Relying Party. who stores it in conjunction with the present user’s account. The authenticator maps the private key to the Relying Party’s RP ID and stores it. Subsequently, only that Relying Party, as identified by its RP ID, is able to employ the scoped credential in authentication ceremonies, via the getAssertion() method. The Relying Party uses its copy of the stored public key to verify the resultant WebAuthn Assertion.

User Consent

User consent means the user agrees with what they are being asked, i.e., it encompasses reading and understanding prompts. User verification encompasses the means employed by the user to indicate consent.

User Verification

The process by which an authenticator locally authorizes the invocation of the authenticatorMakeCredential and authenticatorGetAssertion operations, for example through a touch plus pin code, a password, a gesture (e.g., presenting a fingerprint), or other modality. Note that invocation of said operations implies use of key material managed by the authenticator.

WebAuthn Assertion

The cryptographically signed WebAuthnAssertion object returned by an authenticator as the result of a authenticatorGetAssertion operation.

WebAuthn Client

See Conforming User Agent.

4. Web Authentication API

This section normatively specifies the API for creating and using scoped credentials. Support for deleting credentials is deliberately omitted; this is expected to be done through platform-specific user interfaces rather than from a script. The basic idea is that the credentials belong to the user and are managed by an authenticator, with which the Relying Party interacts through the client (consisting of the browser and underlying OS platform). Scripts can (with the user’s consent) request the browser to create a new credential for future use by the Relying Party. Scripts can also request the user’s permission to perform authentication operations with an existing credential. All such operations are performed in the authenticator and are mediated by the browser and/or platform on the user’s behalf. At no point does the script get access to the credentials themselves; it only gets information about the credentials in the form of objects.

The security properties of this API are provided by the client and the authenticator working together. The authenticator, which holds and manages credentials, ensures that all operations are scoped to a particular web origin, and cannot be replayed against a different origin, by incorporating the origin in its responses. Specifically, as defined in §5.2 Signature Format, the full origin of the requester is included, and signed over, in the attestation statement produced when a new credential is created as well as in all assertions produced by WebAuthn credentials.

Additionally, to maintain user privacy and prevent malicious Relying Parties from probing for the presence of credentials belonging to other Relying Parties, each credential is also associated with a Relying Party Identifier, or RP ID. This RP ID is provided by the client to the authenticator for all operations, and the authenticator ensures that credentials created by a Relying Party can only be used in operations requested by the same RP ID. Separating the origin from the RP ID in this way allows the API to be used in cases where a single Relying Party maintains multiple web origins.

The client facilitates these security measures by providing correct web origins and RP IDs to the authenticator for each operation. Since this is an integral part of the WebAuthn security model, user agents SHOULD only expose this API to callers in secure contexts, as defined in [secure-contexts].

The API is defined by the following Web IDL fragment.

partial interface Navigator {
    readonly attribute WebAuthentication authentication;
};

interface WebAuthentication {
    Promise < ScopedCredentialInfo > makeCredential (
        Account                                 accountInformation,
        sequence < ScopedCredentialParameters > cryptoParameters,
        BufferSource                            attestationChallenge,
        optional CredentialOptions              options
    );

    Promise < WebAuthnAssertion > getAssertion (
        BufferSource               assertionChallenge,
        optional AssertionOptions  options
    );
};

interface ScopedCredentialInfo {
    readonly attribute Credential           credential;
    readonly attribute CryptoKey            publicKey;
    readonly attribute WebAuthnAttestation  attestation;
};

dictionary Account {
    required DOMString rpDisplayName;
    required DOMString displayName;
    DOMString          name;
    DOMString          id;
    DOMString          imageURL;
};

dictionary ScopedCredentialParameters {
    required CredentialType        type;
    required AlgorithmIdentifier   algorithm;
};

dictionary CredentialOptions {
    unsigned long            timeoutSeconds;
    sequence < Credential >  excludeList;
    WebAuthnExtensions       extensions;
};

interface WebAuthnAssertion {
    readonly attribute Credential  credential;
    readonly attribute ArrayBuffer clientData;
    readonly attribute ArrayBuffer authenticatorData;
    readonly attribute ArrayBuffer signature;
};

dictionary AssertionOptions {
    unsigned long           timeoutSeconds;
    sequence < Credential > allowList;
    WebAuthnExtensions      extensions;
};

dictionary WebAuthnExtensions {
};

interface WebAuthnAttestation {
    readonly    attribute DOMString     type;
    readonly    attribute ArrayBuffer   clientData;
    readonly    attribute any           statement;
};

enum CredentialType {
    "ScopedCred"
};

interface Credential {
    readonly attribute CredentialType type;
    readonly attribute BufferSource   id;
};

4.1. WebAuthentication Interface

This interface has two methods, which are described in the following subsections.

4.1.1. Create a new credential (makeCredential() method)

With this method, a script can request the User Agent to create a new credential of a given type and persist it to the underlying platform, which may involve data storage managed by the browser or the OS. The user agent will prompt the user to approve this operation. On success, the promise will be resolved with a ScopedCredentialInfo object describing the newly created credential.

This method takes the following parameters:

• The accountInformation parameter specifies information about the user account for which the credential is being created. This is meant for later use by the authenticator when it needs to prompt the user to select a credential.

• The cryptoParameters parameter supplies information about the desired properties of the credential to be created. The sequence is ordered from most preferred to least preferred. The platform makes a best effort to create the most preferred credential that it can.

• The attestationChallenge parameter contains a challenge intended to be used for generating the attestation statement of the newly created credential.

• The optional options parameter specifies additional options, as described in §4.5 Additional options for Credential Generation (dictionary CredentialOptions).

When this method is invoked, the user agent MUST execute the following algorithm:

1. If timeoutSeconds was specified, check if its value lies within a reasonable range as defined by the platform and if not, correct it to the closest value lying within that range. Set adjustedTimeout to this adjusted value. If timeoutSeconds was not specified then set adjustedTimeout to a platform-specific default.

2. Let promise be a new Promise. Return promise and start a timer for adjustedTimeout seconds. Then asynchronously continue executing the following steps.

3. Set callerOrigin to the origin of the caller. Derive the RP ID from callerOrigin by computing the "public suffix + 1" or "PS+1" (which is also referred to as the "Effective Top-Level Domain plus One" or "eTLD+1") part of callerOrigin [PSL]. Set rpIdHash to the SHA-256 hash of the UTF-8 encoding of the lowercase form of this RP ID.

4. Initialize issuedRequests to an empty list.

5. Process each element of cryptoParameters using the following steps, to produce a new sequence normalizedParameters:

   • Let current be the currently selected element of cryptoParameters.

   • If current.type does not contain a CredentialType supported by this implementation, then stop processing current and move on to the next element in cryptoParameters.

   • Let normalizedAlgorithm be the result of normalizing an algorithm using the procedure defined in [WebCryptoAPI], with alg set to current.algorithm and op set to 'generateKey'. If an error occurs during this procedure, then stop processing current and move on to the next element in cryptoParameters.

   • Add a new object of type ScopedCredentialParameters to normalizedParameters, with type set to current.type and algorithm set to normalizedAlgorithm.

6. If excludeList is undefined, set it to the empty list.

7. If extensions was specified, process any extensions supported by this client platform, to produce the extension data that needs to be sent to the authenticator. Call this data clientExtensions.

8. For each authenticator currently available on this platform: asynchronously invoke the authenticatorMakeCredential operation on that authenticator with callerOrigin, rpIdHash, accountInformation, normalizedParameters, excludeList, attestationChallenge and clientExtensions as parameters. Add a corresponding entry to issuedRequests.

9. While issuedRequests is not empty, perform the following actions depending upon the adjustedTimeout timer and responses from the authenticators:

   • If the adjustedTimeout timer expires, then for each entry in issuedRequests invoke the authenticatorCancel operation on that authenticator and remove its entry from the list.

   • If any authenticator returns a status indicating that the user cancelled the operation, delete that authenticator’s entry from issuedRequests. For each remaining entry in issuedRequests invoke the authenticatorCancel operation on that authenticator and remove its entry from the list.

   • If any authenticator returns an error status, delete the corresponding entry from issuedRequests.

   • If any authenticator indicates success, create a new ScopedCredentialInfo object named value and populate its fields with the values returned from the authenticator. Resolve promise with value and terminate this algorithm.

10. Resolve promise with a DOMException whose name is "NotFoundError", and terminate this algorithm.

During the above process, the user agent SHOULD show some UI to the user to guide them in the process of selecting and authorizing an authenticator.

4.1.2. Use an existing credential (getAssertion() method)

This method is used to discover and use an existing scoped credential, with the user’s consent. The script optionally specifies some criteria to indicate what credentials are acceptable to it. The user agent and/or platform locates credentials matching the specified criteria, and guides the user to pick one that the script should be allowed to use. The user may choose not to provide a credential even if one is present, for example to maintain privacy.

This method takes the following parameters:

• The assertionChallenge parameter contains a challenge that the selected authenticator is expected to sign to produce the assertion.

• The optional options parameter specifies additional options, as described in §4.7 Additional options for Assertion Generation (dictionary AssertionOptions).

When this method is invoked, the user agent MUST execute the following algorithm:

1. If timeoutSeconds was specified, check if its value lies within a reasonable range as defined by the platform and if not, correct it to the closest value lying within that range. Set adjustedTimeout to this adjusted value. If timeoutSeconds was not specified then set adjustedTimeout to a platform-specific default.

2. Let promise be a new Promise. Return promise and start a timer for adjustedTimeout seconds. Then asynchronously continue executing the following steps.

3. Set callerOrigin to the origin of the caller. Derive the RP ID from callerOrigin by computing the "public suffix + 1" or "PS+1" (which is also referred to as the "Effective Top-Level Domain plus One" or "eTLD+1") part of callerOrigin [PSL]. Set rpIdHash to the SHA-256 hash of the UTF-8 encoding of the lowercase form of this RP ID.

4. Initialize issuedRequests to an empty list.

5. If extensions was specified, process any extensions supported by this client platform, to produce the extension data that needs to be sent to the authenticator. Call this data clientExtensions.

6. For each authenticator currently available on this platform, perform the following steps:

   • If allowList is undefined or empty, let credentialList be a list containing a single wildcard entry. Otherwise, execute a platform-specific procedure to determine which of these credentials might be present on this authenticator, and set credentialList to this filtered list.

   • If credentialList is empty, ignore this authenticator and do not perform any of the following per-authenticator steps.

   • Asynchronously invoke the authenticatorGetAssertion operation on this authenticator with callerOrigin, rpIdHash, assertionChallenge, credentialList, and clientExtensions as parameters.

   • Add an entry to issuedRequests, corresponding to this request.

7. While issuedRequests is not empty, perform the following actions depending upon the adjustedTimeout timer and responses from the authenticators:

   • If the timer for adjustedTimeout expires, then for each entry in issuedRequests invoke the authenticatorCancel operation on that authenticator and remove its entry from the list.

   • If any authenticator returns a status indicating that the user cancelled the operation, delete that authenticator’s entry from issuedRequests. For each remaining entry in issuedRequests invoke the authenticatorCancel operation on that authenticator, and remove its entry from the list.

   • If any authenticator returns an error status, delete the corresponding entry from issuedRequests.

   • If any authenticator returns success, create a new WebAuthnAssertion object named value and populate its fields with the values returned from the authenticator. Resolve promise with value and terminate this algorithm.

8. Resolve promise with a DOMException whose name is "NotFoundError", and terminate this algorithm.

During the above process, the user agent SHOULD show some UI to the user to guide them in the process of selecting and authorizing an authenticator with which to complete the operation.

4.2. ScopedCredentialInfo Interface

4.3. User Account Information (dictionary Account)

4.4. Parameters for Credential Generation (dictionary ScopedCredentialParameters)

4.5. Additional options for Credential Generation (dictionary CredentialOptions)

4.6. WebAuthn Assertion (interface WebAuthnAssertion)

Scoped credentials produce a cryptographic signature that provides proof of possession of a private key as well as evidence of user consent to a specific transaction. The structure of these signatures is defined as follows.

4.7. Additional options for Assertion Generation (dictionary AssertionOptions)

4.8. WebAuthn Assertion Extensions (dictionary WebAuthnExtensions)

This is a dictionary containing zero or more extensions as defined in §7 WebAuthn Extensions. An extension is an additional parameter that can be passed to the getAssertion() method and triggers some additional processing by the client platform and/or the authenticator.

If the caller wants to pass extensions to the platform, it SHOULD do so by adding one entry per extension to this dictionary with the extension identifier as the key, and the extension’s value as the value (see §5.2 Signature Format for details).

4.9. Credential Attestation Statement (interface WebAuthnAttestation)

Authenticators also provide some form of attestation. The basic requirement is that the authenticator can produce, for each credential public key, attestation information that can be verified by a Relying Party. Typically, this information contains a signature by an attesting key over the attested public key and a challenge, as well as a certificate or similar information providing provenance information for the attesting key, enabling a trust decision to be made.

This attestation statement is delivered to the Relying Party by the Relying Party’s script running on the client, using methods outside the scope of this specification. It contains all the information that the Relying Party’s server requires to validate the statement, as well as to decode and validate the bindings of both the client and authenticator data.

4.10. Supporting Data Structures

The scoped credential type uses certain data structures that are specified in supporting specifications. These are as follows.

4.10.1. Credential Type enumeration (enum CredentialType)

4.10.2. Unique Identifier for Credential (interface Credential)

This interface contains the attributes that are returned to the caller when a new credential is created, and can be used later by the caller to select a credential for use.

4.10.3. Cryptographic Algorithm Identifier (type AlgorithmIdentifier)

A string or dictionary identifying a cryptographic algorithm and optionally a set of parameters for that algorithm. This type is defined in [WebCryptoAPI].

5. WebAuthn Authenticator model

The API defined in this specification implies a specific abstract functional model for an authenticator. This section describes the authenticator model. Client platforms may implement and expose this abstract model in any way desired. However, the behavior of the client’s Web Authentication API implementation, when operating on the authenticators supported by that platform, MUST be indistinguishable from the behavior specified in §4 Web Authentication API.

In this abstract model, each authenticator stores some number of scoped credentials. Each scoped credential has an identifier which is unique (or extremely unlikely to be duplicated) among all scoped credentials. Each credential is also associated with a Relying Party, whose identity is represented by a Relying Party Identifier (RP ID).

5.1. Authenticator operations

A client must connect to an authenticator in order to invoke any of the operations of that authenticator. This connection defines an authenticator session. An authenticator must maintain isolation between sessions. It may do this by only allowing one session to exist at any particular time, or by providing more complicated session management.

The following operations can be invoked by the client in an authenticator session.

5.1.1. The authenticatorMakeCredential operation

This operation must be invoked in an authenticator session which has no other operations in progress. It takes the following input parameters:

• The web origin of the script on whose behalf the operation is being initiated, as determined by the user agent and the client.

• The SHA-256 hash of the RP ID corresponding to the above web origin, as determined by the user agent and the client.

• The Account information provided by the Relying Party.

• The CredentialType requested by the Relying Party.

• The cryptographic parameters requested by the Relying Party, with the cryptographic algorithms normalized as per the procedure in Web Cryptography API §algorithm-normalization-normalize-an-algorithm.

• A list of Credential objects provided by the Relying Party with the intention that, if any of these are known to the authenticator, it should not create a new credential.

• A challenge provided by the Relying Party to assure freshness of the attestation statement of the new credential.

• Extension data created by the client based on the extensions requested by the Relying Party.

When this operation is invoked, the authenticator obtains user consent for creating a new credential. The prompt for obtaining this consent is shown by the authenticator if it has its own output capability, or by the user agent otherwise. Once user consent is obtained, the authenticator generates the appropriate cryptographic keys and creates a new credential. It also generates an identifier for the credential, such that this identifier is globally unique with high probability across all credentials with the same type across all authenticators. It then associates the credential with the specified RP ID hash such that it will be able to retrieve the RP ID hash later, given the credential ID. Finally, it generates an attestation statement that describes its own attributes as well as some attributes of the credential. For more details on attestation, see §5.3 Credential Attestation Statements.

On successful completion of this operation, the authenticator returns the following to the client:

• The type and unique identifier of the new credential.

• The public key associated with the new credential.

• The attestation statement, including information about the attestation format used.

If the user refuses consent, the authenticator returns an appropriate error status to the client.

5.1.2. The authenticatorGetAssertion operation

This operation must be invoked in an authenticator session which has no other operations in progress. It takes the following input parameters:

• The web origin of the script on whose behalf the operation is being initiated, as determined by the user agent and the client.

• The RP ID hash corresponding to the above web origin, as determined by the user agent and the client.

• A challenge provided by the Relying Party to assure freshness of the assertion produced.

• A list of credentials acceptable to the Relying Party (possibly filtered by the client).

• Extension data created by the client based on the extensions requested by the Relying Party.

When this method is invoked, the authenticator allows the user to select a credential from among the credentials associated with that Relying Party (as determined by an exact match of the RP ID hash) and matching the specified criteria, then obtains user consent for using that credential. The prompt for obtaining this consent may be shown by the authenticator if it has its own output capability, or by the user agent otherwise. Once a credential is selected and user consent is obtained, the authenticator computes a cryptographic signature using the credential’s private key and constructs an assertion signature as specified in §5.2 Signature Format.

On successful completion, the authenticator returns to the user agent:

• The identifier of the credential used to generate the signature.

• The authenticatorData used to generate the signature.

• The signature itself.

If the authenticator cannot find any credential corresponding to the specified Relying Party that matches the specified criteria, it terminates the operation and returns an error.

If the user refuses consent, the authenticator returns an appropriate error status to the client.

5.1.3. The authenticatorCancel operation

This operation takes no input parameters and returns no result.

When this operation is invoked by the client in an authenticator session, it has the effect of terminating any authenticatorMakeCredential or authenticatorGetAssertion operation currently in progress in that authenticator session. The authenticator stops prompting for, or accepting, any user input related to authorizing the canceled operation. The client ignores any further responses from the authenticator for the canceled operation.

This operation is ignored if it is invoked in an authenticator session which does not have an authenticatorMakeCredential or authenticatorGetAssertion operation currently in progress.

5.2. Signature Format

WebAuthn signatures are bound to various contextual data. These data are observed, and added at different levels of the stack as a signature request passes from the server to the authenticator. In verifying a signature, the server checks these bindings against expected values.

The components of a system using WebAuthn can be divided into three layers:

1. The Relying Party (RP), which uses the WebAuthn services. The RP consists of a server component and a web-application running in a browser.

2. The WebAuthn Client platform, which consists of the User Agent and the OS and device on which it executes.

3. The Authenticator itself, which provides key management and cryptographic signatures. This may be embedded in the WebAuthn client, or housed in a separate device entirely. In the latter case, the interface between the WebAuthn client and the authenticator is a separately-defined protocol.

This specification defines the common signature format shared by all the above layers. This includes how the different contextual bindings are encoded, signed over, and delivered to the RP.

The goals of this design can be summarized as follows.

• The scheme for generating signatures should accommodate cases where the link between the client platform and authenticator is very limited, in bandwidth and/or latency. Examples include Bluetooth Low Energy and Near-Field Communication.

• The data processed by the authenticator should be small and easy to interpret in low-level code. In particular, authenticators should not have to parse high-level encodings such as JSON.

• Both the client platform and the authenticator should have the flexibility to add contextual bindings as needed.

• The design aims to reuse as much as possible of existing encoding formats in order to aid adoption and implementation.

The contextual bindings are divided in two: Those added by the RP or the client platform, referred to as client data; and those added by the authenticator, referred to as the authenticator data. The client data must be signed over, but an authenticator is otherwise not interested in its contents. To save bandwidth and processing requirements on the authenticator, the client platform hashes the client data and sends only the result to the authenticator. The authenticator signs over the combination of this hash, and its own authenticator data.

5.2.1. Client data used in WebAuthn signatures (dictionary ClientData)

The client data represents the contextual bindings of both the Relying Party and the client platform. It is a key-value mapping with string-valued keys. Values may be any type that has a valid encoding in JSON. Its structure is defined by the following Web IDL.

dictionary ClientData {
    required DOMString           challenge;
    required DOMString           facet;
    required DOMString           rpId;
    required AlgorithmIdentifier hashAlg;
    JsonWebKey                   tokenBinding;
    WebAuthnExtensions           extensions;
};

5.2.2. Authenticator data

The authenticator data encodes contextual bindings made by the authenticator itself. The authenticator data has a compact but extensible encoding. This is desired since authenticators can be devices with limited capabilities and low power requirements, with much simpler software stacks than the client platform components.

The encoding of authenticator data is a byte array of 37 bytes or more, as follows.

Note that the RP ID hash is originally received from the client when the credential is created, and again when an assertion is generated. However, it differs from other client data in some important ways. First, unlike the client data, the RP ID of a credential does not change between operations but instead remains the same for the lifetime of that credential. Secondly, it is validated by the authenticator during the authenticatorGetAssertion operation, by making sure that the RP ID hash associated with the requested credential exactly matches the RP ID hash supplied by the client. These differences also explain why the RP ID hash is always a SHA-256 hash instead of being crypto-agile like the clientDataHash; for a given RP ID, we need the hash to be computed the same way by all clients for all operations so that authenticators can move between clients without losing interoperability.

The TUP flag SHALL be set if and only if the authenticator detected a user through an authenticator specific gesture. The RFU bits in the flags byte SHALL be set to zero.

If the authenticator does not include any extension data, it MUST set the ED flag in the first byte to zero, and to one if extension data is included.

The figure below shows a visual representation of the authenticator data structure.

Note: The authenticatorData describes its own length: If the ED flag is not set, it is always 37 bytes long. If the ED flag is set, it is 37 bytes plus the CBOR map that follows.

5.2.3. Generating a signature

Before making a request to an authenticator, the client platform layer SHALL perform the following steps.

1. Represent the parameters passed in by the RP in the form of a ClientData structure.

2. Let clientDataJSON be the UTF-8 encoded JSON serialization [RFC7159] of this ClientData dictionary.

3. Let clientDataHash be the hash (computed using hashAlg) of clientDataJSON, as an array.

The clientDataHash value is delivered to the authenticator.

The hash algorithm hashAlg used to compute clientDataHash is included in the ClientData object. This way it is available to the Relying Party and it is also hashed over when computing clientDataHash and hence anchored in the signature itself.

A raw cryptographic signature must assert the integrity of both the client data and the authenticator data. Thus, an authenticator SHALL compute a signature over the concatenation of the authenticatorData and the clientDataHash.

Note: A simple, undelimited concatenation is safe to use here because the authenticatorData describes its own length. The clientDataHash (which potentially has a variable length) is always the last element.

The authenticator MUST return both the authenticatorData and the raw signature back to the client. The client, in turn, MUST return clientDataJSON, authenticatorData and the signature to the RP. The clientDataJSON is returned in the clientData member of the WebAuthnAssertion and AttestationStatement structures.

5.3. Credential Attestation Statements

An attestation statement is a specific type of signature, which contains statements about a credential itself and the authenticator that holds it. Therefore, the procedures for generating attestation statements closely parallel those for generating WebAuthn assertions as described in §5.2 Signature Format, though the semantics of the contextual bindings are quite different.

This specification defines a number of attestation formats, i.e., ways to serialize the data being attested to by the Authenticator. The reason is to be able to support existing devices like TPMs and other platform-specific formats. Each attestation format provides the ability to cryptographically attest to a public key, the authenticator model, and contextual data to a remote party. They differ in the details of how the attestation statement is laid out, and how its components are computed. The different attestation formats are defined in §6 Defined Attestation Formats.

This specification also defines a number of attestation types. These define how a Relying Party establishes trust in a particular attestation statement, after verifying that it is cryptographically valid.

Attestation formats are orthogonal to attestation types, i.e. attestation formats in general are not restricted to a single attestation type. For example the "packed" attestation format (see below) can be used in conjunction with all attestation types. Broadly speaking, attestation formats pertain to the formatting / encoding of the attestation statement rawData, while attestation types pertain to the semantics. The privacy, the security and the operational characteristic of attestation mainly depends on the attestation type while it is largely independent from the attestation format.

5.3.1. Attestation Types

WebAuthn supports multiple attestation types:

Full Basic Attestation

In the case of full basic attestation [UAFProtocol], the Authenticator’s attestation private key is specific to an Authenticator model. That means that an Authenticator of the same model typically shares the same attestation private key. This model is also used for FIDO UAF 1.0 and FIDO U2F 1.0.

Surrogate Basic Attestation

In the case of surrogate basic attestation [UAFProtocol], the Authenticator doesn’t have any specific attestation key. Instead it uses the authentication key to (self-)sign the (surrogate) attestation message. Authenticators without meaningful protection measures for an attestation private key typically use this attestation type.

Privacy CA

In this case, the Authenticator owns an authenticator-specific (endorsement) key. This key is used to securely communicate with a trusted third party, the Privacy CA. The Authenticator can generate multiple attestation key pairs and asks the Privacy CA to issue an attestation certificate for it. Using this approach, the Authenticator can limit the exposure of the endorsement key (which is a global correlation handle) to Privacy CA(s). Attestation keys can be requested for each scoped credential individually.

Note: This concept typically leads to multiple attestation certificates. The attestation certificate requested most recently is called "active".

Direct Anonymous Attestation (DAA)

In this case, the Authenticator receives DAA credentials from a single DAA-Issuer. These DAA credentials are used along with blinding to sign the attestation data. The concept of blinding avoids the DAA credentials being misused as global correlation handle. WebAuthn supports DAA using elliptic curve cryptography and bilinear pairings, called ECDAA (see [FIDOEcdaaAlgorithm]) in this specification.

Compliant servers MUST support all attestation types. Authenticators can choose what attestation type to implement.

Note: Relying Parties can always decide what attestation types are acceptable to them by policy.

5.3.2. Verifying an Attestation Statement

This section outlines the recommended algorithm for verifying an attestation statement, independent of attestation format.

Upon receiving an attestation statement, the Relying Party shall:

1. Verify that the attestation statement is properly formatted.

2. If alg is not ECDAA (e.g., not "ED256" and not "ED512"):

   • Look up the attestation root certificate from a trusted source. The FIDO Metadata Service [FIDOMetadataService] provides an easy way to access such information. The claimedAAGUID can be used for this lookup.

   • Verify that the attestation certificate chain is valid and chains up to a trusted root (following [RFC5280]).

   • Verify that the attestation certificate has the right Extended Key Usage (EKU) based on the WebAuthn Authenticator type (as denoted by the type member). In case of a type="tpm", this EKU shall be OID "2.23.133.8.3". Defined attestation formats are in §6 Defined Attestation Formats.

   • If the attestation format is "android", verify that the attestation certificate is issued to the hostname "attest.android.com" (see SafetyNet online documentation).

   • Verify that all issuing CA certificates in the chain are valid and not revoked.

   • Verify the signature on rawData using the attestation certificate public key and algorithm as identified by alg.

   • Verify rawData syntax and that it doesn’t contradict the signing algorithm specified in alg.

   • If the attestation certificate contains an extension with OID 1 3 6 1 4 1 45724 1 1 4 (id-fido-gen-ce-aaguid) verify that the value of this extension matches claimedAAGUID. This identifies the Authenticator model.

   • If such extension doesn’t exist, the attestation root certificate is dedicated to a single Authenticator model.

3. If alg is ECDAA (e.g., "ED256", "ED512"):

   • Look up the DAA root key from a trusted source. The FIDO Metadata Service [FIDOMetadataService] provides an easy way to access such information. The claimedAAGUID can be used for this lookup.

   • Perform DAA-Verify on signature for rawData (see [FIDOEcdaaAlgorithm]).

   • Verify rawData syntax and that it doesn’t contradict the signing algorithm specified in alg.

   • The DAA root key is dedicated to a single Authenticator model.

4. Verify the contextual bindings (e.g., channel binding) from the clientData (see §5.2.3 Generating a signature).

5. Verify that user verification method and other authenticator characteristics related to this authenticator model, match the Relying Party policy. The FIDO Metadata Service [FIDOMetadataService] provides an easy way to access the authenticator characteristics.

The Relying Party MAY take any of the below actions when verification of an attestation statement fails:

• Reject the request, such as a registration request, associated with the attestation statement.

• Accept the request associated with the attestation statement but treat the attested Scoped Credential as one with surrogate basic attestation (see §5.3.1 Attestation Types), if policy allows it. If doing so, there is no cryptographic proof that the Scoped Credential has been generated by a particular Authenticator model. See [FIDOSecRef] and [UAFProtocol] for more details on the relevance on attestation.

Verification of attestation statements requires that the relying party trusts the root of the attestation certificate chain. Also, the Relying Party must have access to certificate status information for the intermediate CA certificates. The relying party must also be able to build the attestation certificate chain if the client didn’t provide this chain in the attestation information.

5.3.3. Security Considerations

5.3.3.1. Privacy

Attestation keys may be used to track users or link various online identities of the same user together. This may be mitigated in several ways, including:

• A WebAuthn Authenticator manufacturer may choose to ship all of their devices with the same (or a fixed number of) attestation key(s) (called Full Basic Attestation). This will anonymize the user at the risk of not being able to revoke a particular attestation key should its WebAuthn Authenticator be compromised.

• A WebAuthn Authenticator may be capable of dynamically generating different attestation keys (and requesting related certificates) per origin (following the Privacy CA approach). For example, a WebAuthn Authenticator can ship with a master attestation key (and certificate), and combined with a cloud operated privacy CA, can dynamically generate per origin attestation keys and attestation certificates.

• A WebAuthn Authenticator can implement direct anonymous attestation (see [FIDOEcdaaAlgorithm]). Using this scheme, the authenticator generates a blinded attestation signature. This allows the Relying Party to verify the signature using the DAA root key, but the attestation signature doesn’t serve as a global correlation handle.

5.3.3.2. Attestation Certificate and Attestation Certificate CA Compromise

When an intermediate CA or a root CA used for issuing attestation certificates is compromised, WebAuthn Authenticator attestation keys are still safe although their certificates can no longer be trusted. A WebAuthn Authenticator manufacturer that has recorded the public attestation keys for their devices can issue new attestation certificates for these keys from a new intermediate CA or from a new root CA. If the root CA changes, the Relying Parties must update their trusted root certificates accordingly.

A WebAuthn Authenticator attestation certificate must be revoked by the issuing CA if its key has been compromised. A WebAuthn Authenticator manufacturer may need to ship a firmware update and inject new attestation keys and certificates into already manufactured WebAuthn Authenticators, if the exposure was due to a firmware flaw. (The process by which this happens is out of scope for this specification.) No further valid attestation statements can be made by the affected WebAuthn Authenticators unless the WebAuthn Authenticator manufacturer has this capability.

If attestation certificate validation fails due to a revoked intermediate attestation CA certificate, and the Relying Party’s policy requires rejecting the registration/authentication request in these situations, then it is recommended that the Relying Party also un-registers (or marks as "surrogate attestation" (see §5.3.1 Attestation Types), policy permitting) scoped credentials that were registered post the CA compromise date using an attestation certificate chaining up to the same intermediate CA. It is thus recommended that Relying Parties remember intermediate attestation CA certificates during Authenticator registration in order to un-register related Scoped Credentials if the registration was performed after revocation of such certificates.

If a DAA attestation key has been compromised, it can be added to the RogueList (i.e., the list of revoked authenticators) maintained by the related DAA-Issuer. The Relying Party should verify whether an authenticator belongs to the RogueList when performing DAA-Verify. The FIDO Metadata Service [FIDOMetadataService] provides an easy way to access such information.

5.3.3.3. Attestation Certificate Hierarchy

A 3 tier hierarchy for attestation certificates is recommended (i.e., Attestation Root, Attestation Issuing CA, Attestation Certificate). It is also recommended that for each WebAuthn Authenticator device line (i.e., model), a separate issuing CA is used to help facilitate isolating problems with a specific version of a device.

If the attestation root certificate is not dedicated to a single WebAuthn Authenticator device line (i.e., AAGUID), the AAGUID must be specified either in the attestation certificate itself or it must be specified in rawData.

6. Defined Attestation Formats

WebAuthn supports pluggable attestation data formats. This allows support of TPM generated attestation data as well as support of other WebAuthn authenticators. As mentioned in §5.3 Credential Attestation Statements, these differ in how the attestation statement is computed and formatted. This section defines these details.

The contents of the attestation data must be controlled (i.e., generated or at least verified) by the authenticator itself.

6.1. Packed Attestation ("packed")

Packed attestation is a WebAuthn optimized format of attestation data. It uses a very compact but still extensible encoding method. Encoding this format can even be implemented by authenticators with very limited resources (e.g., secure elements).

A Packed Attestation statement has the following format:

interface AttestationStatement {
    readonly    attribute unsigned long version;
    readonly    attribute ArrayBuffer   claimedAAGUID;
    readonly    attribute DOMString[]   x5c;
    readonly    attribute DOMString     alg;
    readonly    attribute ArrayBuffer   rawData;
    readonly    attribute ArrayBuffer   signature;
};

6.1.1. Attestation rawData

The attestation data encodes contextual bindings made by the authenticator itself. Unlike client data, the authenticator data has a compact but extensible encoding. This is desired since authenticators can be devices with limited capabilities and low power requirements, with much simpler software stacks than the client platform components.

The field rawData for this type is a byte array of 45 bytes + length of public key + length of KeyHandle + potentially more extensions. The first bytes before the extensions have a fixed layout as follows:

The TUP flag SHALL be set if and only if the authenticator detected a user through an authenticator-specific gesture. The RFU bits in the flags byte SHALL be cleared (i.e., zeroed).

If the authenticator does not wish to add extensions, it MUST clear the ED flag in the third byte.

6.1.2. Signature

The signature is computed over the rawData field. The following algorithms must be implemented by servers:

1. "ES256" [RFC7518]

2. "RS256" [RFC7518]

3. "PS256" [RFC7518]

4. "ED256" [FIDOEcdaaAlgorithm]

Authenticators can choose which algorithm(s) to implement. Relying Parties must implement all the algorithms implemented by the authenticators that they support.

6.1.3. Packed attestation statement certificate requirements

Note: In the case of DAA attestation [FIDOEcdaaAlgorithm] no attestation certificate is used.

The attestation certificate MUST have the following fields/extensions:

• Version must be set to 3.

• Subject field MUST be set to:

  Subject-C

  Country where the Authenticator vendor is incorporated

  Subject-O

  Legal name of the Authenticator vendor

  Subject-OU

  Authenticator Attestation

  Subject-CN

  No stipulation.

• If the related attestation root certificate is used for multiple authenticator models, the Extension OID 1 3 6 1 4 1 45724 1 1 4 (id-fido-gen-ce-aaguid) MUST be present, containing the AAGUID as value.

• The Basic Constraints extension MUST have the CA component set to false

• An Authority Information Access (AIA) extension with entry id-ad-ocsp and a CRL Distribution Point extension [RFC5280] are both optional as the status of attestation certificates is available through the FIDO Metadata Service [FIDOMetadataService].

6.2. TPM Attestation ("tpm")

This attestation format returns an attestation statement in the same format as defined in §6.1 Packed Attestation ("packed"). However the rawData and signature fields are computed differently, as described below.

6.2.1. Attestation rawData

The value of rawData is either a TPM_CERTIFY_INFO or a TPM_CERTIFY_INFO2 structure [TPMv1-2-Part2] sections 11.1 and 11.2, if version equals 1. Else, if version equals 2, it MUST be a TPMS_ATTEST structure as defined in [TPMv2-Part2] section 10.12.9.

The field "extraData" (in the case of TPMS_ATTEST) or the field "data" (in the case of TPM_CERTIFY_INFO or TPM_CERTIFY_INFO2) MUST contain the clientDataHash (see §5.2 Signature Format).

6.2.2. Signature

If version equals 1, (i.e., for TPM 1.2), RSASSA-PKCS1-v1_5 signature algorithm (section 8.2 of [RFC3447]) can be used by WebAuthn Authenticators (i.e., alg="RS256").

If version equals 2, the following algorithms can be used by WebAuthn Authenticators:

1. TPM_ALG_RSASSA (0x14). This is the same algorithm RSASSA-PKCS1-v1_5 as for version 1 but for use with TPMv2. alg="RS256".

2. TPM_ALG_RSAPSS (0x16); alg="PS256".

3. TPM_ALG_ECDSA (0x18); alg="ES256".

4. TPM_ALG_ECDAA (0x1A); alg="ED256".

5. TPM_ALG_SM2 (0x1B); alg="SM256".

Relying Parties must implement all the algorithms implemented by the authenticators that they support.

The signature is computed over the rawData field.

6.2.3. TPM attestation statement certificate requirements

TPM attestation certificate MUST have the following fields/extensions:

• Version must be set to 3.

• Subject field MUST be set to empty.

• The Subject Alternative Name extension must be set as defined in [TPMv2-EK-Profile] section 3.2.9 if "version" equals 2 and [TPMv1-2-Credential-Profiles] section 3.2.9 if "version" equals 1.

• The Extended Key Usage extension MUST contain the "joint-iso-itu-t(2) internationalorganizations(23) 133 tcg-kp(8) tcg-kp-AIKCertificate(3)" OID.

• The Basic Constraints extension MUST have the CA component set to false

• An Authority Information Access (AIA) extension with entry id-ad-ocsp and a CRL Distribution Point extension [RFC5280] are both optional as the status of attestation certificates is available through the FIDO Metadata Service [FIDOMetadataService].

6.3. Android Attestation (type="android")

When the Authenticator in question is a platform-provided Authenticator on the Android platform, the attestation statement is based on the SafetyNet API.

This type of attestation statement is formatted as follows:

interface AndroidAttestation {
    readonly attribute unsigned long version;
    readonly attribute DOMString     safetyNetResponse;
};

6.3.1. Signature

For this attestation format, the ClientData dictionary is extended in the following way:

dictionary AndroidAttestationClientData : ClientData {
    JsonWebKey    publicKey;
    boolean       isInsideSecureHardware;
    DOMString     userAuthentication;
    unsigned long userAuthenticationValidityDurationSeconds; // optional
};

In order to generate an attestation statement, the client MUST create clientDataJSON by UTF8-encoding a structure of type AndroidAttestationClientData, and compute clientDataHash as the hash of clientDataJSON. It must then provide clientDataHash as the Nonce value when requesting the SafetyNet attestation.

6.3.2. Verifying AndroidClientData specific contextual bindings

A Relying Party shall verify the clientData contextual bindings (see step 4 in §5.3.2 Verifying an Attestation Statement) as follows:

• Check that AndroidAttestationClientData.challenge equals the attestationChallenge that was passed into the makeCredential() call.

• Check that the facet and tokenBinding parameters in the AndroidAttestationClientData match the Relying Party App.

• Check that AndroidAttestationClientData.publicKey is the same key as the one returned in the ScopedCredentialInfo by the makeCredential call.

• Check that the hash of the clientDataJSON matches the nonce attribute in the payload of the safetynetResponse JWS.

• Check that the ctsProfileMatch attribute in the payload of the safetynetResponse is true.

• Check that the apkPackageName attribute in the payload of the safetynetResponse matches package name of the application calling SafetyNet API.

• Check that the apkDigestSha256 attribute in the payload of the safetynetResponse matches the package hash of the application calling SafetyNet API.

• Check that the apkCertificateDigestSha256 attribute in the payload of the safetynetResponse matches the hash of the signing certificate of the application calling SafetyNet API.

7. WebAuthn Extensions

The mechanism for generating scoped credentials, as well as requesting and generating WebAuthn assertions, as defined in §4 Web Authentication API, can be extended to suit particular use cases. Each case is addressed by defining a registration extension and/or a authentication extension. Extensions can define additions to the following steps and data:

• makeCredential() request parameters for registration extension.

• getAssertion() request parameters for authentication extensions.

• Client processing, and the ClientData structure, for registration extensions and authentication extensions.

• Authenticator processing, and the authenticatorData structure, for registration extensions and authentication extensions.

When requesting an assertion for a scoped credential, a Relying Party can list a set of extensions to be used, if they are supported by the client and/or the authenticator. It sends the client arguments for each extension in the getAssertion() call (for authentication extensions) or makeCredential() call (for registration extensions) to the client platform. The client platform performs additional processing for each extension that it supports, and augments ClientData as required by the extension. In addition, the client collects the authenticator arguments for the above extensions, and passes them to the authenticator in the authenticatorMakeCredential call (for registration extensions) or authenticatorGetAssertion call (for authentication extensions). These authenticator arguments are passed as name-value pairs, with the extension identifier as the name, and the corresponding authenticator argument as the value. The authenticator, in turn, performs additional processing for the extensions that it supports, and augments authenticatorData as specified by the extension.

All WebAuthn extensions are optional for both clients and authenticators. Thus, any extensions requested by a Relying Party may be ignored by the client browser or OS and not passed to the authenticator at all, or they may be ignored by the authenticator. Ignoring an extension is never considered a failure in WebAuthn API processing, so when Relying Parties include extensions with any API calls, they must be prepared to handle cases where some or all of those extensions are ignored.

Clients wishing to support the widest possible range of extensions may choose to pass through any extensions that they do not recognize to authenticators, generating the authenticator argument by simply encoding the client argument in CBOR. All WebAuthn extensions MUST be defined in such a way that this implementation choice does not endanger the user’s security or privacy. For instance, if an extension requires client processing, it could be defined in a manner that ensures such a naïve pass-through will produce a semantically invalid authenticator argument, resulting in the extension being ignored by the authenticator. Since all extensions are optional, this will not cause a functional failure in the API operation.

7.1. Extension identifiers

Extensions are identified by a string, chosen by the extension author. Extension identifiers should aim to be globally unique, e.g., by including the defining entity such as myCompany_extension.

Extensions that may exist in multiple versions should take care to include a version in their identifier. In effect, different versions are thus treated as different extensions, e.g., myCompany_extension_01

Extensions defined in this specification use a fixed prefix of webauthn for the extension identifiers. This prefix should not be used for extensions not defined by the W3C.

7.2. Defining extensions

A definition of an extension must specify, at minimum, an extension identifier and an extension client argument sent via the getAssertion() or makeCredential() call. Additionally, extensions may specify additional values in ClientData, authenticatorData (in the case of authentication extensions), or both. Finally, if the extension requires any authenticator processing, it must also specify an authenticator argument to be sent via the authenticatorGetAssertion or authenticatorMakeCredential call.

Any extension that requires client processing MUST specify a method of augmenting ClientData that unambiguously lets the Relying Party know that the extension was honored by the client. Similarly, any extension that requires authenticator processing MUST specify a method of augmenting authenticatorData to let the Relying Party know that the extension was honored by the authenticator.

7.3. Extending request parameters

An extension defines up to two request arguments. The client argument is passed from the Relying Party to the client in the getAssertion() or makeCredential() call, while the authenticator argument is passed from the client to the authenticator during the processing of these calls.

A Relying Party simultaneously requests the use of an extension and sets its client argument by including an entry in the extensions option to the makeCredential() or getAssertion() call. The entry key MUST be the extension identifier, and the value MUST be the client argument.

var assertionPromise = credentials.getAssertion(..., /* extensions */ {
    "webauthnExample_foobar": 42
});

Extension definitions MUST specify the valid values for their client argument. Clients SHOULD ignore extensions with an invalid client argument. If an extension does not require any parameters from the Relying Party, it SHOULD be defined as taking a Boolean client argument, set to true to signify that the extension is requested by the Relying Party.

Extensions that only affect client processing need not specify an authenticator argument. Extensions that affect authenticator processing MUST specify a method of computing the authenticator argument from the client argument. For extensions that do not require additional parameters, and are defined as taking a Boolean client argument set to true, this method SHOULD consist of passing an authenticator argument of true (CBOR major type 7, value 21).

Note: Extensions should aim to define authenticator arguments that are as small as possible. Some authenticators communicate over low-bandwidth links such as Bluetooth Low-Energy or NFC.

7.4. Extending client processing

Extensions may define additional processing requirements on the client platform during the creation of credentials or the generation of an assertion. In order for the Relying Party to verify the processing took place, or if the processing has a result value that the Relying Party needs to be aware of, the extension should specify a client data value to be included in the ClientData structure.

The client data value may be any value that can be encoded using JSON. If any extension processed by a client defines such a value, the client SHOULD include a dictionary in ClientData with the key extensions. For each such extension, the client SHOULD add an entry to this dictionary with the extension identifier as the key, and the extension’s client data value.

Extensions that require authenticator processing MUST define the process by which the client argument can be used to determine the authenticator argument.

7.5. Extending authenticator processing

Extensions that define additional authenticator processing may similarly define an authenticator data value. The value may be any data that can be encoded in CBOR. An authenticator that processes a authentication extension that defines such a value must include it in the authenticatorData.

As specified in §5.2.2 Authenticator data, the authenticator data value of each processed extension is included in the extended data part of the authenticatorData. This part is a CBOR map, with extension identifiers as keys, and the authenticator data value of each extension as the value.

7.6. Example extension

This section is not normative.

To illustrate the requirements above, consider a hypothetical extension "Geo". This extension, if supported, lets both clients and authenticators embed their geolocation in assertions.

The extension identifier is chosen as webauthnExample_geo. The client argument is the constant value true, since the extension does not require the Relying Party to pass any particular information to the client, other than that it requests the use of the extension. The Relying Party sets this value in its request for an assertion:

var assertionPromise =
    credentials.getAssertion("SGFuIFNvbG8gc2hvdCBmaXJzdC4",
        {}, /* Empty filter */
        { 'webauthnExample_geo': true });

The extension defines the additional client data to be the client’s location, if known, as a GeoJSON [GeoJSON] point. The client constructs the following client data:

{
    ...,
    'extensions': {
        'webauthnExample_geo': {
            'type': 'Point',
            'coordinates': [65.059962, -13.993041]
        }
    }
}

The extension also requires the client to set the authenticator parameter to the fixed value true.

Finally, the extension requires the authenticator to specify its geolocation in the authenticator data, if known. The extension e.g. specifies that the location shall be encoded as a two-element array of floating point numbers, encoded with CBOR. An authenticator does this by including it in the authenticatorData. As an example, authenticator data may be as follows (notation taken from [RFC7049]):

81 (hex)                                    -- Flags, ED and TUP both set.
20 05 58 1F                                 -- Signature counter
A1                                          -- CBOR map of one element
    73                                      -- Key 1: CBOR text string of 19 bytes
        77 65 62 61 75 74 68 6E 45 78 61
        6D 70 6C 65 5F 67 65 6F             -- "webauthnExample_geo" UTF-8 string
    82                                      -- Value 1: CBOR array of two elements
        FA 42 82 1E B3                      -- Element 1: Latitude as CBOR encoded float
        FA C1 5F E3 7F                      -- Element 2: Longitude as CBOR encoded float

8. Pre-defined extensions

This section defines an initial set of extensions. These are recommended for implementation by user agents targeting broad interoperability.

8.1. Transaction authorization

This authentication extension allows for a simple form of transaction authorization. A Relying Party can specify a prompt string, intended for display on a trusted device on the authenticator.

Extension identifier

webauthn_txAuthSimple

Client argument

A single UTF-8 encoded string prompt.

Client processing

None, except default forwarding of client argument to authenticator argument.

Authenticator argument

The client argument encoded as a CBOR text string (major type 3).

Authenticator processing

The authenticator MUST display the prompt to the user before performing the user verification / test of user presence. The authenticator may insert line breaks if needed.

Authenticator data

A single UTF-8 string, representing the prompt as displayed (including any eventual line breaks).

The generic version of this extension allows images to be used as prompts as well. This allows authenticators without a font rendering engine to be used and also supports a richer visual appearance.

Extension identifier

webauthn_txAuthGeneric

Client argument

A CBOR map with one pair of data items (CBOR tagged as 0xa1). The pair of data items consists of

1. one UTF-8 encoded string contentType, containing the MIME-Type of the content, e.g. "image/png"

2. and the content itself, encoded as CBOR byte array.

Client processing

None, except default forwarding of client argument to authenticator argument.

Authenticator argument

The client argument encoded as a CBOR map.

Authenticator processing

The authenticator MUST display the content to the user before performing the user verification / test of user presence. The authenticator may add other information below the content. No changes are allowed to the content itself, i.e., inside content boundary box.

Authenticator data

The hash value of the content which was displayed. The authenticator MUST use the same hash algorithm as it uses for the signature itself.

8.2. Authenticator Selection Extension

This registration extension allows a Relying Party to guide the selection of the authenticator that will be leveraged when creating the credential. It is intended primarily for Relying Parties that wish to tightly control the experience around credential creation.

Extension identifier

webauthn_authnSel

Client argument

A sequence of AAGUIDs:

typedef sequence < AAGUID > AuthenticatorSelectionList;

Each AAGUID corresponds to an authenticator model that is acceptable to the Relying Party for this credential creation. The list is ordered by decreasing preference.

An AAGUID is defined as an array containing the globally unique identifier of the authenticator model being sought.

typedef BufferSource AAGUID;

Client processing

This extension can only be used during makeCredential(). If the client supports the Authenticator Selection Extension, it MUST use the first available authenticator whose AAGUID is present in the AuthenticatorSelectionList. If none of the available authenticators match a provided AAGUID, the client MUST select an authenticator from among the available authenticators to generate the credential. If an authenticator was selected from AuthenticatorSelectionList, its AAGUID MUST be added by the client to the ClientData as the client data value for this extension.

Authenticator argument

There is no authenticator argument.

Authenticator processing

None.

8.3. SupportedExtensions Extension

Extension identifier

webauthn_exts

Client argument

The Boolean value true to indicate that this extension is requested by the Relying Party.

Client processing

None, except default forwarding of client argument to authenticator argument.

Authenticator argument

The Boolean value true, encoded in CBOR (major type 7, value 21).

Authenticator processing

The authenticator augments the authenticator data with a list of extensions that the authenticator supports, as defined below. This extension can be added to attestation statements.

Authenticator data

The SupportedExtensions extension is a list (CBOR array) of extension identifiers encoded as UTF-8 Strings.

8.4. User Verification Index (UVI) Extension

Extension identifier

webauthn_uvi

Client argument

The Boolean value true to indicate that this extension is requested by the Relying Party.

Client processing

None, except default forwarding of client argument to authenticator argument.

Authenticator argument

The Boolean value true, encoded in CBOR (major type 7, value 21).

Authenticator processing

The authenticator augments the authenticator data with a user verification index indicating the method used by the user to authorize the operation, as defined below. This extension can be added to attestation statements and assertions.

Authenticator data

The user verification index (UVI) is a value uniquely identifying a user verification data record. The UVI is encoded as CBOR byte string (type 0x58). Each UVI value MUST be specific to the related key (in order to provide unlinkability). It also must contain sufficient entropy that makes guessing impractical. UVI values MUST NOT be reused by the Authenticator (for other biometric data or users).

The UVI data can be used by servers to understand whether an authentication was authorized by the exact same biometric data as the initial key generation. This allows the detection and prevention of "friendly fraud".

As an example, the UVI could be computed as SHA256(KeyID | SHA256(rawUVI)), where the rawUVI reflects (a) the biometric reference data, (b) the related OS level user ID and (c) an identifier which changes whenever a factory reset is performed for the device, e.g. rawUVI = biometricReferenceData | OSLevelUserID | FactoryResetCounter.

Servers supporting UVI extensions MUST support a length of up to 32 bytes for the UVI value.

Example for rawData containing one UVI extension

F1 D0                                       -- This is a WebAuthn packed rawData object
81                                          -- TUP and ED set
00 00 00 01                                 -- (initial) signature counter
...                                         -- all public key alg etc.
A1                                          -- extension: CBOR map of one element
    6C                                      -- Key 1: CBOR text string of 11 bytes
        77 65 62 61 75 74 68 6E 5F 75 76 69 -- "webauthn_uvi" UTF-8 string
    58 20                                   -- Value 1: CBOR byte string with 0x20 bytes
        00 43 B8 E3 BE 27 95 8C             -- the UVI value itself
        28 D5 74 BF 46 8A 85 CF
        46 9A 14 F0 E5 16 69 31
        DA 4B CF FF C1 BB 11 32
        82

9. IANA Considerations

This specification registers the algorithm names "S256", "S384", "S512", and "SM3" with the IANA JSON Web Algorithms registry as defined in section "Cryptographic Algorithms for Digital Signatures and MACs" in [RFC7518].

These names follow the naming strategy in draft-ietf-oauth-spop-15.

10. Sample scenarios

This section is not normative.

In this section, we walk through some events in the lifecycle of a scoped credential, along with the corresponding sample code for using this API. Note that this is an example flow, and does not limit the scope of how the API can be used.

As was the case in earlier sections, this flow focuses on a use case involving an external first-factor authenticator with its own display. One example of such an authenticator would be a smart phone. Other authenticator types are also supported by this API, subject to implementation by the platform. For instance, this flow also works without modification for the case of an authenticator that is embedded in the client platform. The flow also works for the case of an authenticator without its own display (similar to a smart card) subject to specific implementation considerations. Specifically, the client platform needs to display any prompts that would otherwise be shown by the authenticator, and the authenticator needs to allow the client platform to enumerate all the authenticator’s credentials so that the client can have information to show appropriate prompts.

10.1. Registration

This is the first time flow, when a new credential is created and registered with the server.

1. The user visits example.com, which serves up a script. At this point, the user must already be logged in using a legacy username and password, or additional authenticator, or other means acceptable to the Relying Party.

2. The Relying Party script runs the code snippet below.

3. The client platform searches for and locates the authenticator.

4. The client platform connects to the authenticator, performing any pairing actions if necessary.

5. The authenticator shows appropriate UI for the user to select the authenticator on which the new credential will be created, and obtains a biometric or other authorization gesture from the user.

6. The authenticator returns a response to the client platform, which in turn returns a response to the Relying Party script. If the user declined to select an authenticator or provide authorization, an appropriate error is returned.

7. If a new credential was created,

   • The Relying Party script sends the newly generated public key to the server, along with additional information about public key such as attestation that it is held in trusted hardware.

   • The server stores the public key in its database and associates it with the user as well as with the strength of authentication indicated by attestation, also storing a friendly name for later use.

   • The script may store data such as the credential ID in local storage, to improve future UX by narrowing the choice of credential for the user.

The sample code for generating and registering a new key follows:

var webauthnAPI = navigator.authentication;

if (!webauthnAPI) { /* Platform not capable. Handle error. */ }

var userAccountInformation = {
    rpDisplayName: "Acme",
    displayName: "John P. Smith",
    name: "johnpsmith@example.com",
    id: "1098237235409872",
    imageURL: "https://pics.acme.com/00/p/aBjjjpqPb.png"
};

// This Relying Party will accept either an ES256 or RS256 credential, but
// prefers an ES256 credential.
var cryptoParams = [
    {
        type: "ScopedCred",
        algorithm: "ES256"
    },
    {
        type: "ScopedCred",
        algorithm: "RS256"
    }
];
var challenge = "Y2xpbWIgYSBtb3VudGFpbg";
var options = { timeoutSeconds: 300,  // 5 minutes
                excludeList: [],      // No excludeList
                extensions: {"webauthn.location": true}  // Include location information
                                               // in attestation
};

// Note: The following call will cause the authenticator to display UI.
webauthnAPI.makeCredential(userAccountInformation, cryptoParams, challenge, options)
    .then(function (newCredentialInfo) {
    // Send new credential info to server for verification and registration.
}).catch(function (err) {
    // No acceptable authenticator or user refused consent. Handle appropriately.
});

10.2. Authentication

This is the flow when a user with an already registered credential visits a website and wants to authenticate using the credential.

1. The user visits example.com, which serves up a script.

2. The script asks the client platform for a WebAuthn identity assertion, providing as much information as possible to narrow the choice of acceptable credentials for the user. This may be obtained from the data that was stored locally after registration, or by other means such as prompting the user for a username.

3. The Relying Party script runs one of the code snippets below.

4. The client platform searches for and locates the authenticator.

5. The client platform connects to the authenticator, performing any pairing actions if necessary.

6. The authenticator presents the user with a notification that their attention is required. On opening the notification, the user is shown a friendly selection menu of acceptable credentials using the account information provided when creating the credentials, along with some information on the origin that is requesting these keys.

7. The authenticator obtains a biometric or other authorization gesture from the user.

8. The authenticator returns a response to the client platform, which in turn returns a response to the Relying Party script. If the user declined to select a credential or provide an authorization, an appropriate error is returned.

9. If an assertion was successfully generated and returned,

   • The script sends the assertion to the server.

   • The server examines the assertion and validates that it was correctly generated. If so, it looks up the identity associated with the associated public key; that identity is now authenticated. If the public key is not recognized by the server (e.g., deregistered by server due to inactivity) then the authentication has failed; each Relying Party will handle this in its own way.

   • The server now does whatever it would otherwise do upon successful authentication -- return a success page, set authentication cookies, etc.

If the Relying Party script does not have any hints available (e.g., from locally stored data) to help it narrow the list of credentials, then the sample code for performing such an authentication might look like this:

var webauthnAPI = navigator.authentication;

if (!webauthnAPI) { /* Platform not capable. Handle error. */ }

var challenge = "Y2xpbWIgYSBtb3VudGFpbg";
var options = {
                timeoutSeconds = 300,  // 5 minutes
                allowList: [{ type: "ScopedCred" }]
              };

webauthnAPI.getAssertion(challenge, options)
    .then(function (assertion) {
    // Send assertion to server for verification
}).catch(function (err) {
    // No acceptable credential or user refused consent. Handle appropriately.
});

On the other hand, if the Relying Party script has some hints to help it narrow the list of credentials, then the sample code for performing such an authentication might look like the following. Note that this sample also demonstrates how to use the extension for transaction authorization.

var webauthnAPI = navigator.authentication;

if (!webauthnAPI) { /* Platform not capable. Handle error. */ }

var challenge = "Y2xpbWIgYSBtb3VudGFpbg";
var acceptableCredential1 = {
    type: "ScopedCred",
    id: "ISEhISEhIWhpIHRoZXJlISEhISEhIQo="
};
var acceptableCredential2 = {
    type: "ScopedCred",
    id: "cm9zZXMgYXJlIHJlZCwgdmlvbGV0cyBhcmUgYmx1ZQo="
};
    
var options = {
                timeoutSeconds: 300,  // 5 minutes
                allowList: [acceptableCredential1, acceptableCredential2];
                extensions: { 'webauthn.txauth.simple':
                   "Wave your hands in the air like you just don’t care" };
              };

webauthnAPI.getAssertion(challenge, options)
    .then(function (assertion) {
    // Send assertion to server for verification
}).catch(function (err) {
    // No acceptable credential or user refused consent. Handle appropriately.
});

10.3. Decommissioning

The following are possible situations in which decommissioning a credential might be desired. Note that all of these are handled on the server side and do not need support from the API specified here.

• Possibility #1 -- user reports the credential as lost.

  • User goes to server.example.net, authenticates and follows a link to report a lost/stolen device.

  • Server returns a page showing the list of registered credentials with friendly names as configured during registration.

  • User selects a credential and the server deletes it from its database.

  • In future, the Relying Party script does not specify this credential in any list of acceptable credentials, and assertions signed by this credential are rejected.

• Possibility #2 -- server deregisters the credential due to inactivity.

  • Server deletes credential from its database during maintenance activity.

  • In the future, the Relying Party script does not specify this credential in any list of acceptable credentials, and assertions signed by this credential are rejected.

• Possibility #3 -- user deletes the credential from the device.

  • User employs a device-specific method (e.g., device settings UI) to delete a credential from their device.

  • From this point on, this credential will not appear in any selection prompts, and no assertions can be generated with it.

  • Sometime later, the server deregisters this credential due to inactivity.

11. Acknowledgements