1.1. Requirements Notation and Conventions

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶

All applications of JSON Web Signature (JWS) data structures [JWS] and JSON Web Encryption (JWE) data structures [JWE] as discussed in this specification utilize the JWS Compact Serialization or the JWE Compact Serialization; the JWS JSON Serialization and the JWE JSON Serialization are not used. Choosing a single serialization is intended to facilitate interoperability.¶

1.2. Terminology

This specification uses the terms "access token", "authorization code", "authorization server", "client", "client authentication", "client identifier", "protected resource", and "resource server" defined by OAuth 2.0 [RFC6749], and the terms "Claim Name" and "JSON Web Token (JWT)" defined by "JSON Web Token (JWT)" [JWT].¶

This specification defines the following term:¶

Resource Identifier:

The protected resource's resource identifier, which is a URL that uses the https scheme and has no fragment component. As specified in Section 2 of [RFC8707], it also SHOULD NOT include a query component, but it is recognized that there are cases that make a query component a useful and necessary part of a resource identifier. Protected resource metadata is published at a .well-known location [RFC8615] derived from this resource identifier, as described in Section 3.¶

2.1. Human-Readable Resource Metadata

Human-readable resource metadata values and resource metadata values that reference human-readable content MAY be represented in multiple languages and scripts. For example, the values of fields such as resource_name, resource_documentation, resource_tos_uri, and resource_policy_uri might have multiple locale-specific metadata values to facilitate use in different locations.¶

To specify the languages and scripts, language tags [BCP47] are added to resource metadata parameter names, delimited by a # character. Since member names as discussed in JSON [RFC8259] are case sensitive, it is RECOMMENDED that language tag values used in Claim Names be spelled using the character case with which they are registered in the "Language Subtag Registry" [IANA.Language]. In particular, normally, language names are spelled with lowercase characters, region names are spelled with uppercase characters, and languages are spelled with mixed-case characters. However, since language tag values are case insensitive per [BCP47], implementations SHOULD interpret the language tag values supplied in a case-insensitive manner. Per the recommendations in [BCP47], language tag values used in metadata parameter names should only be as specific as is necessary. For instance, using fr might be sufficient in many contexts, rather than fr-CA or fr-FR.¶

For example, a resource could represent its name in English as "resource_name#en": "My Resource" and its name in Italian as "resource_name#it": "La mia bella risorsa" within its metadata. Any or all of these names MAY be displayed to the end user, choosing which names to display based on system configuration, user preferences, or other factors.¶

If any human-readable field is sent without a language tag, parties using it MUST NOT make any assumptions about the language, character set, or script of the string value, and the string value MUST be used as is wherever it is presented in a user interface. To facilitate interoperability, it is RECOMMENDED that each kind of human-readable metadata provided include an instance of its metadata parameter without any language tags in addition to any language-specific parameters, and it is RECOMMENDED that any human-readable fields sent without language tags contain values suitable for display on a wide variety of systems.¶

2.2. Signed Protected Resource Metadata

In addition to JSON elements, metadata values MAY also be provided as a signed_metadata value, which is a JSON Web Token (JWT) [JWT] that asserts metadata values about the protected resource as a bundle. A set of metadata parameters that can be used in signed metadata as claims are defined in Section 2. The signed metadata MUST be digitally signed or MACed (protected with a Message Authentication Code) using a JSON Web Signature (JWS) [JWS] and MUST contain an iss (issuer) claim denoting the party attesting to the claims in the signed metadata. Consumers of the metadata MAY ignore the signed metadata if they do not support this feature. If the consumer of the metadata supports signed metadata, metadata values conveyed in the signed metadata MUST take precedence over the corresponding values conveyed using plain JSON elements.¶

Signed metadata is included in the protected resource metadata JSON object using this OPTIONAL metadata parameter:¶

signed_metadata

A JWT containing metadata parameters about the protected resource as claims. This is a string value consisting of the entire signed JWT. A signed_metadata parameter SHOULD NOT appear as a claim in the JWT; it is RECOMMENDED to reject any metadata in which this occurs.¶

3.1. Protected Resource Metadata Request

A protected resource metadata document MUST be queried using an HTTP GET request at the previously specified URL.¶

The consumer of the metadata would make the following request when the resource identifier is https://resource.example.com and the well-known URI path suffix is oauth-protected-resource to obtain the metadata, since the resource identifier contains no path component:¶

  GET /.well-known/oauth-protected-resource HTTP/1.1
  Host: resource.example.com

If the resource identifier value contains a path or query component, any terminating slash (/) following the host component MUST be removed before inserting /.well-known/ and the well-known URI path suffix between the host component and the path and/or query components. The consumer of the metadata would make the following request when the resource identifier is https://resource.example.com/resource1 and the well-known URI path suffix is oauth-protected-resource to obtain the metadata, since the resource identifier contains a path component:¶

  GET /.well-known/oauth-protected-resource/resource1 HTTP/1.1
  Host: resource.example.com

Using path components enables supporting multiple resources per host. This is required in some multi-tenant hosting configurations. This use of .well-known is for supporting multiple resources per host; unlike its use in [RFC8615], it does not provide general information about the host.¶

3.2. Protected Resource Metadata Response

The response is a set of metadata parameters about the protected resource's configuration. A successful response MUST use the 200 OK HTTP status code and return a JSON object using the application/json content type that contains a set of metadata parameters as its members that are a subset of the metadata parameters defined in Section 2. Additional metadata parameters MAY be defined and used; any metadata parameters that are not understood MUST be ignored.¶

Parameters with multiple values are represented as JSON arrays. Parameters with zero values MUST be omitted from the response.¶

An error response uses the applicable HTTP status code value.¶

The following is a non-normative example response:¶

  HTTP/1.1 200 OK
  Content-Type: application/json

  {
   "resource":
     "https://resource.example.com",
   "authorization_servers":
     ["https://as1.example.com",
      "https://as2.example.net"],
   "bearer_methods_supported":
     ["header", "body"],
   "scopes_supported":
     ["profile", "email", "phone"],
   "resource_documentation":
     "https://resource.example.com/resource_documentation.html"
  }

3.3. Protected Resource Metadata Validation

The resource value returned MUST be identical to the protected resource's resource identifier value into which the well-known URI path suffix was inserted to create the URL used to retrieve the metadata. If these values are not identical, the data contained in the response MUST NOT be used.¶

If the protected resource metadata was retrieved from a URL returned by the protected resource via the WWW-Authenticate resource_metadata parameter, then the resource value returned MUST be identical to the URL that the client used to make the request to the resource server. If these values are not identical, the data contained in the response MUST NOT be used.¶

These validation actions can thwart impersonation attacks, as described in Section 7.3.¶

The recipient MUST validate that any signed metadata was signed by a key belonging to the issuer and that the signature is valid. If the signature does not validate or the issuer is not trusted, the recipient SHOULD treat this as an error condition.¶

5.1. WWW-Authenticate Response

This specification introduces a new parameter in the WWW-Authenticate HTTP response header field to indicate the protected resource metadata URL:¶

resource_metadata:

The URL of the protected resource metadata.¶

The response below is an example of a WWW-Authenticate header that includes the resource identifier.¶

HTTP/1.1 401 Unauthorized
WWW-Authenticate: Bearer resource_metadata=
  "https://resource.example.com/.well-known/oauth-protected-resource"

The HTTP status code in the example response above is defined by [RFC6750].¶

This parameter MAY also be used in WWW-Authenticate responses using authorization schemes other than "Bearer" [RFC6750], such as the DPoP scheme defined by [RFC9449].¶

The resource_metadata parameter MAY be combined with other parameters defined in other extensions, such as the max_age parameter defined by [RFC9470].¶

5.2. Changes to Resource Metadata

At any point, for any reason determined by the resource server, the protected resource MAY respond with a new WWW-Authenticate challenge that includes a value for the protected resource metadata URL to indicate that its metadata may have changed. If the client receives such a WWW-Authenticate response, it SHOULD retrieve the updated protected resource metadata and use the new metadata values obtained, after validating them as described in Section 3.3. Among other things, this enables a resource server to change which authorization servers it uses without any other coordination with clients.¶

5.3. Client Identifier and Client Authentication

The way in which the client identifier is established at the authorization server is out of scope for this specification.¶

This specification is intended to be deployed in scenarios where the client has no prior knowledge about the resource server and where the resource server might or might not have prior knowledge about the client.¶

There are some existing methods by which an unrecognized client can make use of an authorization server, such as using Dynamic Client Registration [RFC7591] to register the client prior to initiating the authorization flow. Future OAuth extensions might define alternatives, such as using URLs to identify clients.¶

5.4. Compatibility with Other Authentication Methods

Resource servers MAY return other WWW-Authenticate headers indicating various authentication schemes. This allows the resource server to support clients that may or may not implement this specification and allows clients to choose their preferred authentication scheme.¶

7.1. TLS Requirements

Implementations MUST support TLS. They MUST follow the guidance in [BCP195], which provides recommendations and requirements for improving the security of deployed services that use TLS.¶

The use of TLS at the protected resource metadata URLs protects against information disclosure and tampering.¶

7.2. Scopes

The scopes_supported parameter is the list of scopes the resource server is willing to disclose that it supports. It is not meant to indicate that an OAuth client should request all scopes in the list. The client SHOULD still follow OAuth best practices and request tokens with as limited a scope as possible for the given operation, as described in Section 2.3 of "Best Current Practice for OAuth 2.0 Security" [RFC9700].¶

7.3. Impersonation Attacks

TLS certificate checking MUST be performed by the client as described in [RFC9525] when making a protected resource metadata request. Checking that the server certificate is valid for the resource identifier URL prevents adversary-in-the-middle and DNS-based attacks. These attacks could cause a client to be tricked into using an attacker's resource server, which would enable impersonation of the legitimate protected resource. If an attacker can accomplish this, they can access the resources that the affected client has access to, using the protected resource that they are impersonating.¶

An attacker may also attempt to impersonate a protected resource by publishing a metadata document that contains a resource metadata parameter using the resource identifier URL of the protected resource being impersonated but that contains information of the attacker's choosing. This would enable it to impersonate that protected resource, if accepted by the client. To prevent this, the client MUST ensure that the resource identifier URL it is using as the prefix for the metadata request exactly matches the value of the resource metadata parameter in the protected resource metadata document received by the client, as described in Section 3.3.¶

7.4. Audience-Restricted Access Tokens

If a client expects to interact with multiple resource servers, the client SHOULD request audience-restricted access tokens using [RFC8707], and the authorization server SHOULD support audience-restricted access tokens.¶

Without audience-restricted access tokens, a malicious resource server (RS1) may be able to use the WWW-Authenticate header to get a client to request an access token with a scope used by a legitimate resource server (RS2), and after the client sends a request to RS1, then RS1 could reuse the access token at RS2.¶

While this attack is not explicitly enabled by this specification and is possible in a plain OAuth 2.0 deployment, it is made somewhat more likely by the use of dynamically configured clients. As such, the use of audience-restricted access tokens and Resource Indicators [RFC8707] is RECOMMENDED when using the features in this specification.¶

7.5. Publishing Metadata in a Standard Format

Publishing information about the protected resource in a standard format makes it easier for both legitimate clients and attackers to use the protected resource. Whether a protected resource publishes its metadata in an ad hoc manner or in the standard format defined by this specification, the same defenses against attacks that might be mounted that use this information should be applied.¶

7.6. Authorization Servers

To support use cases in which the set of legitimate authorization servers to use with the protected resource is enumerable, this specification defines the authorization_servers metadata parameter, which enables explicitly listing them. Note that if the set of legitimate protected resources to use with an authorization server is also enumerable, lists in the protected resource metadata and authorization server metadata should be cross-checked against one another for consistency when these lists are used by the application profile.¶

Secure determination of appropriate authorization servers to use with a protected resource for all use cases is out of scope for this specification. This specification assumes that the client has a means of determining appropriate authorization servers to use with a protected resource and that the client is using the correct metadata for each protected resource. Implementers need to be aware that if an inappropriate authorization server is used by the client, an attacker may be able to act as an adversary-in-the-middle proxy to a valid authorization server without it being detected by the authorization server or the client.¶

The ways to determine the appropriate authorization servers to use with a protected resource are, in general, application dependent. For instance, some protected resources are used with a fixed authorization server or a set of authorization servers, the locations of which may be known via out-of-band mechanisms. Alternatively, as described in this specification, the locations of the authorization servers could be published by the protected resource as metadata values. In other cases, the set of authorization servers that can be used with a protected resource can be dynamically changed by administrative actions or by changes to the set of authorization servers adhering to a trust framework. Many other means of determining appropriate associations between protected resources and authorization servers are also possible.¶

7.7. Server-Side Request Forgery (SSRF)

The OAuth client is expected to fetch the authorization server metadata based on the value of the issuer in the resource server metadata. Since this specification enables clients to interoperate with RSs and ASes it has no prior knowledge of, this opens a risk for Server-Side Request Forgery (SSRF) attacks by malicious users or malicious resource servers. Clients SHOULD take appropriate precautions against SSRF attacks, such as blocking requests to internal IP address ranges. Further recommendations can be found in the Open Worldwide Application Security Project (OWASP) SSRF Prevention Cheat Sheet [OWASP.SSRF].¶

7.8. Phishing

This specification may be deployed in a scenario where the desired HTTP resource is identified by a user-selected URL. If this resource is malicious or compromised, it could mislead the user into revealing their account credentials or authorizing unwanted access to OAuth-controlled capabilities. This risk is reduced, but not eliminated, by following best practices for OAuth user interfaces, such as providing clear notice to the user, displaying the authorization server's domain name, supporting origin-bound phishing-resistant authenticators, supporting the use of password managers, and applying heuristic checks such as domain reputation.¶

7.9. Differences Between Unsigned and Signed Metadata

Unsigned metadata is integrity protected by the use of TLS at the site where it is hosted. This means that its security is dependent upon the Internet Public Key Infrastructure using X.509 (PKIX), as described in [RFC9525]. Signed metadata is additionally integrity protected by the JWS signature applied by the issuer, which is not dependent upon the Internet PKI.¶

When using unsigned metadata, the party issuing the metadata is the protected resource itself, which is represented by the resource value in the metadata, whereas when using signed metadata, the party issuing the metadata is represented by the iss (issuer) claim in the signed metadata. When using signed metadata, applications can make trust decisions based on the issuer that performed the signing -- information that is not available when using unsigned metadata. How these trust decisions are made is out of scope for this specification.¶

7.10. Metadata Caching

Protected resource metadata is retrieved using an HTTP GET request, as specified in Section 3.1. Normal HTTP caching behaviors apply, meaning that the GET request may retrieve a cached copy of the content, rather than the latest copy. Implementations should utilize HTTP caching directives such as Cache-Control with max-age, as defined in [RFC9111], to enable caching of retrieved metadata for appropriate time periods.¶

8.1. OAuth Protected Resource Metadata Registry

This specification establishes the "OAuth Protected Resource Metadata" registry for OAuth 2.0 protected resource metadata names. The registry records the protected resource metadata parameter and a reference to the specification that defines it.¶

8.2. OAuth Authorization Server Metadata Registry

IANA has registered the following authorization server metadata parameter in the "OAuth Authorization Server Metadata" registry established in "OAuth 2.0 Authorization Server Metadata" [RFC8414].¶

8.3. Well-Known URIs Registry

This specification registers the well-known URI defined in Section 3 in the "Well-Known URIs" registry [IANA.well-known].¶