openid@justin.richer.org http://justin.richer.org/

OpenID Heart Working Group The OAuth 2.0 protocol framework defines a mechanism to allow a resource owner to delegate access to a protected resource for a client application. This specification profiles the OAuth 2.0 protocol framework to increase baseline security, provide greater interoperability, and structure deployments in a manner specifically applicable to (but not limited to) the healthcare domain.

This document profiles the OAuth 2.0 web authorization framework for use in the context of securing web-facing application programming interfaces (APIs), particularly Representational State Transfer (RESTful) APIs. The OAuth 2.0 specifications accommodate a wide range of implementations with varying security and usability considerations, across different types of software clients. To achieve this flexibility, the standard makes many security controls optional. OAuth implementations using only the minimum mandatory security measures require minimal effort on the part of developers and users, but they also fail to prevent known attacks and are unsuitable for protecting sensitive data. The OAuth 2.0 client, protected resource, and authorization server profiles defined in this document serve two purposes: Define a mandatory baseline set of security controls suitable for a wide range of use cases, while maintaining reasonable ease of implementation and functionality Identify optional advanced security controls for sensitive use cases where heightened risks justify more stringent controls that increase the required implementation effort and may reduce or restrict functionality This OAuth profile is intended to be shared broadly, and ideally to influence OAuth implementations in other domains besides health care.

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 RFC 2119. All uses of JSON Web Signature (JWS) and JSON Web Encryption (JWE) data structures 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.

This specification uses the terms "Access Token", "Authorization Code", "Authorization Endpoint", "Authorization Grant", "Authorization Server", "Client", "Client Authentication", "Client Identifier", "Client Secret", "Grant Type", "Protected Resource", "Redirection URI", "Refresh Token", "Resource Owner", "Resource Server", "Response Type", and "Token Endpoint" defined by OAuth 2.0, the terms "Claim Name", "Claim Value", and "JSON Web Token (JWT)" defined by JSON Web Token (JWT), and the terms defined by OpenID Connect Core 1.0.

This specification defines requirements for the following components: OAuth 2.0 clients OAuth 2.0 authorization servers OAuth 2.0 protected resources The specification also defines features for interaction between these components: Client to authorization server Protected resource to authorization server When a HEART-compliant component is interacting with other HEART-compliant components, in any valid combination, all components MUST fully conform to the features and requirements of this specification. All interaction with non-HEART components is outside the scope of this specification. A HEART-compliant OAuth 2.0 authorization server MUST support all features as described in this specification. A general-purpose authorization server MAY support additional features for use with non-HEART clients and protected resources. A HEART-compliant OAuth 2.0 client MUST use all functions as described in this specification. A general-purpose client library MAY support additional features for use with non-HEART authorization servers and protected resources. A HEART-compliant OAuth 2.0 protected resource MUST use all functions as described in this specification. A general-purpose protected resource library MAY support additional features for use with non-HEART authorization servers and clients.

Full clients and browser-embedded clients making a request to the authorization endpoint MUST use an unpredictable value for the state parameter with at least 128 bits of entropy. Clients MUST validate the value of the state parameter upon return to the redirect URI and MUST ensure that the state value is securely tied to the user’s current session (e.g., by relating the state value to a session identifier issued by the client software to the browser). Clients MUST include their full redirect URIs in the authorization request. To prevent open redirection and other injection attacks, the authorization server MUST match the entire redirect URI using a direct string comparison against registered values and MUST reject requests with invalid or missing redirect URIs. The following is a sample response from a web-based client to the end user’s browser for the purpose of redirecting the end user to the authorization server's authorization endpoint:

This causes the browser to send the following request to the authorization endpoint:

Full clients and direct access clients as defined above MUST authenticate to the authorization server's token endpoint using a JWT assertion as defined by the JWT Profile for OAuth 2.0 Client Authentication and Authorization Grants and the private_key_jwt method defined in OpenID Connect Core. The assertion MUST use the claims as follows: the client ID of the client creating the token the client ID of the client creating the token the URL of the authorization server's token endpoint the time that the token was created by the client the expiration time, after which the token MUST be considered invalid a unique identifier generated by the client for this authentication. This identifier MUST contain at least 128 bits of entropy and MUST NOT be re-used by any subsequent authentication token. The following sample claim set illustrates the use of the required claims for a client authentication JWT as defined in this profile; additional claims MAY be included in the claim set.

The JWT assertion MUST be signed by the client using the client's private key. See for mechanisms by which the client can make its public key known to the server. The authorization server MUST support the RS256 signature method (the Rivest, Shamir, and Adleman (RSA) signature algorithm with a 256-bit hash) and MAY use other asymmetric signature methods listed in the JSON Web Algorithms (JWA) specification. The following sample JWT contains the above claims and has been signed using the RS256 JWS algorithm and the client's own private key (with line breaks for display purposes only):

This is sent in the request to the token endpoint as in the following example:

The following profile descriptions give patterns of deployment for use in different types of client applications based on the OAuth grant type. The resource owner password credentials grant type defined in is intentionally omitted from this discussion, and MUST NOT be used under these profiles. Additional grant types, such as assertions, chained tokens, or other mechanisms, are out of scope of this profile and must be covered separately by appropriate profile documents.

This client type applies to clients that act on behalf of a particular resource owner and require delegation of that user’s authority to access the protected resource. Furthermore, these clients are capable of interacting with a separate web browser application to facilitate the resource owner's interaction with the authentication endpoint of the authorization server. These clients MUST use the authorization code flow of OAuth 2 by sending the resource owner to the authorization endpoint to obtain authorization. The user MUST authenticate to the authorization endpoint. The user’s web browser is then redirected back to a URI hosted by the client, from which the client can obtain an authorization code passed as a query parameter. The client then presents that authorization code along with its own credentials to the authorization server's token endpoint to obtain an access token. These clients MUST be associated with a unique public key, as described in . This client type MAY request and be issued a refresh token if the security parameters of the access request allow for it.

This client type applies to clients that act on behalf of a particular resource owner and require delegation of that user's authority to access the protected resource. Furthermore, these clients are capable of interacting with a separate web browser application to facilitate the resource owner's interaction with the authentication endpoint of the authorization server. In particular, this client type runs natively on the resource owner's device, often leading to many identical instances of a piece of software operating in different environments and running simultaneously for different end users. These clients MUST use the authorization code flow of OAuth 2 by sending the resource owner to the authorization endpoint to obtain authorization. The user MUST authenticate to the authorization endpoint. The user’s web browser is then redirected back to a URI hosted by the client, from which the client can obtain an authorization code passed as a query parameter. The client then presents that authorization code along to the authorization server's token endpoint to obtain an access token. These clients MUST be associated with a unique public key, as described in . Native clients MUST use dynamic client registration to obtain a separate client id for each instance, and MUST use their client key to protect calls to the token endpoint. Native applications using dynamic registration SHOULD generate a unique public and private key pair on the device and register that public key value with the authorization server. Alternatively, an authorization server MAY issue a public and private key pair to the client as part of the registration process. In such cases, the authorization server MUST discard its copy of the private key. Client credentials MUST NOT be shared among instances of client software. Native applications SHOULD use PKCE, and applications which do use PKCE MUST use the S256 code challenge mechanism. This client type MAY request and be issued a refresh token if the security parameters of the access request allow for it.

This client type applies to clients that act on behalf of a particular resource owner and require delegation of that user’s authority to access the protected resource. Furthermore, these clients are embedded within a web browser and effectively share an active session between systems. These clients use the implicit flow of OAuth 2 by sending a resource owner to the authorization endpoint to obtain authorization. The user MUST authenticate to the authorization endpoint. The user’s web browser is then redirected back to a URI hosted by the client, from which the client can directly obtain an access token. Since the client itself never authenticates to the server and the token is made available directly to the browser, this flow is appropriate only for clients embedded within a web browser, such as a JavaScript client with no back-end server component. Wherever possible, it is preferable to use the authorization code flow due to its superior security properties. This client type MUST NOT request or be issued a refresh token. Access tokens issued to this type of client MUST be short lived and SHOULD expire when the user's authenticated session with the client expires.

This profile applies to clients that connect directly to protected resources and do not act on behalf of a particular resource owner, such as those clients that facilitate bulk transfers. These clients use the client credentials flow of OAuth 2 by sending a request to the token endpoint with the client's credentials and obtaining an access token in the response. Since this profile does not involve an authenticated user, this flow is appropriate only for trusted applications, such as those that would traditionally use a developer key. For example, a partner system that performs bulk data transfers between two systems would be considered a direct access client. This client type MUST NOT request or be issued a refresh token.

All clients MUST register with the authorization server. For client software that may be installed on multiple client instances, such as native applications or web application software, each client instance MUST receive a unique client identifier from the authorization server.

Clients using the authorization code or implicit grant types MUST register their full redirect URIs. The Authorization Server MUST validate the redirect URI given by the client at the authorization endpoint using strict string comparison. A client MUST protect the values passed back to its redirect URI by ensuring that the redirect URI is one of the following: Hosted on a website with Transport Layer Security (TLS) protection (a Hypertext Transfer Protocol – Secure (HTTPS) URI) Hosted on the local domain of the client (e.g., http://localhost/) Hosted on a client-specific non-remote-protocol URI scheme (e.g., myapp://) Clients MUST NOT have URIs in more than one category and SHOULD NOT have multiple redirect URIs on different domains. Clients MUST NOT forward values passed back to their redirect URIs to other arbitrary or user-provided URIs (a practice known as an "open redirector”).

Full clients using the authorization code grant type or direct access clients using the client credentials grant type MUST have a public and private key pair for use in authentication to the token endpoint. These clients MUST register their public keys in their client registration metadata by either sending the public key directly in the jwks field or by registering a jwks_uri that MUST be reachable by the authorization server. It is RECOMMENDED that clients use a jwks_uri if possible as this allows for key rotation more easily. The jwks field or the content available from the jwks_uri of a client MUST contain a public key in JSON Web Key Set (JWK Set) format. The authorization server MUST validate the content of the client's registered jwks_uri document and verify that it contains a JWK Set. The following example is of a 2048-bit RSA key:

For reference, the corresponding public/private key pair for this public key is the following (in JWK format):

Note that the second example contains both the public and private keys, while the first example contains the public key only. Native client applications MAY omit their key during registration if they are a public client using PKCE.

Clients SHOULD send bearer tokens passed in the Authentication header as defined by . Clients MAY use the form-parameter or query-parameter methods in . Authorized requests MUST be made over TLS, and clients MUST validate the protected resource server's certificate. An example of an OAuth-protected call to the OpenID Connect UserInfo endpoint, sending the token in the Authorization header, follows:

All servers MUST conform to applicable recommendations found in the Security Considerations sections of and those found in the OAuth Threat Model Document. The authorization server MUST protect all communications to and from its OAuth endpoints using TLS.

The authorization server MUST support the authorization_code, implicit, and client_credentials grant types as described in . The authorization server MUST limit each registered client (identified by a client ID) to a single grant type only, since a single piece of software will be functioning at runtime in only one of the modes described in . Clients that have multiple modes of operation MUST have a separate client ID for each mode.

The authorization server MUST enforce client authentication as described above for the authorization code and client credentials grant types. The authorization server MUST validate all redirect URIs for authorization code and implicit grant types.

Authorization servers MUST support dynamic client registration, and clients MAY register using the Dynamic Client Registration Protocol for authorization code or implicit grant types. Clients MUST NOT dynamically register for the client credentials grant type. Authorization servers MAY limit the scopes available to dynamically registered clients. Authorization servers MUST signal to end users that a client was dynamically registered on the authorization screen. Authorization servers MAY accept signed software statements as described in issued to client software developers from a trusted registration entity. The software statement can be used to tie together many instances of the same client software that will be run, dynamically registered, and authorized separately at runtime. The software statement MUST include the following client metadata parameters: array of redirect URIs used by the client; subject to the requirements listed in grant type used by the client; must be "authorization_code” or "implicit” client's public key in JWK Set format; if jwks_uri is used it MUST be reachable by the Authorization Server and point to the client's public key set human-readable name of the client URL of a web page containing further information about the client

When prompting the end user with an interactive approval page, the authorization server MUST indicate to the user: Whether the client was dynamically registered, or else statically registered by a trusted administrator Whether the client is associated with a software statement, and in which case provide information about the trusted issuer of the software statement What kind of access the client is requesting (including scope, target resource, etc.)

The authorization server MUST provide an OpenID Connect service discovery endpoint listing the components relevant to the OAuth protocol: The fully qualified issuer URL of the server The fully qualified URL of the server's authorization endpoint defined by OAuth 2.0 The fully qualified URL of the server's token endpoint defined by OAuth 2.0 The fully qualified URL of the server's introspection endpoint defined by OAuth Token Introspection The fully qualified URL of the server's revocation endpoint defined by OAuth 2.0 Token Revocation The fully qualified URI of the server's public key in JWK Set format If the authorization server is also an OpenID Connect Provider, it MUST provide a discovery endpoint meeting the requirements listed in Section 3.5 of the HEART OpenID Connect profile. The following example shows the JSON document found at a discovery endpoint for an authorization server:

Clients and protected resources SHOULD cache this discovery information. It is RECOMMENDED that servers provide cache information through HTTP headers and make the cache valid for at least one week. The server MUST provide its public key in JWK Set format. The key MUST contain the following fields: The key ID of the key pair used to sign this token The key type The default algorithm used for this key The following is an example of a 2048-bit RSA public key:

Clients and protected resources SHOULD cache this key. It is RECOMMENDED that servers provide cache information through HTTP headers and make the cache valid for at least one week.

Token revocation allows a client to signal to an authorization server that a given token will no longer be used. An authorization server MUST revoke the token if the client requesting the revocation is the client to which the token was issued, the client has permission to revoke tokens, and the token is revocable. A client MUST immediately discard the token and not use it again after revoking it.

An authorization server MUST support the Proof Key for Code Exchange (PKCE) extension to the authorization code flow, including support for the S256 code challenge method. The authorization server MUST NOT allow a HEART client to use the plain code challenge method.

The authorization server MUST compare a client's registered redirect URIs with the redirect URI presented during an authorization request using an exact string match.

Unlike the core OAuth protocol, the HEART profile intends to allow compliant protected resources to connect to compliant authorization servers.

In order to facilitate interoperability with multiple protected resources, all HEART-compliant authorization servers issue cryptographically signed tokens in the JSON Web Token (JWT) format. The information carried in the JWT is intended to allow a protected resource to quickly test the integrity of the token without additional network calls, and to allow the protected resource to determine which authorization server issued the token. When combined with discovery, this information is sufficient to programmatically locate the token introspection service, which is in turn used for conveying additional security information about the token. The server MUST issue tokens as JWTs with, at minimum, the following claims: The issuer URL of the server that issued the token The client id of the client to whom this token was issued The expiration time (integer number of seconds since from 1970-01-01T00:00:00Z UTC), after which the token MUST be considered invalid A unique JWT Token ID value with at least 128 bits of entropy. This value MUST NOT be re-used in another token. Clients MUST check for reuse of jti values and reject all tokens issued with duplicate jti values. The server MAY issue tokens with additional fields, including the following as defined here: The identifier of the end-user that authorized this client, or the client id of a client acting on its own behalf (such as a bulk transfer). Since this information could potentially leak private user information, it should be used only when needed. The audience of the token, an array containing the identifier(s) of protected resource(s) for which the token is valid, if this information is known. The aud claim may contain multiple values if the token is valid for multiple protected resources. Note that at runtime, the authorization server may not know the identifiers of all possible protected resources at which a token may be used. The following sample claim set illustrates the use of the required claims for an access token as defined in this profile; additional claims MAY be included in the claim set:

The access tokens MUST be signed with JWS. The authorization server MUST support the RS256 signature method for tokens and MAY use other asymmetric signing methods as defined in the IANA JSON Web Signatures and Encryption Algorithms registry. The JWS header MUST contain the following fields: The key ID of the key pair used to sign this token This example access token has been signed with the server's private key using RS256:

Refresh tokens SHOULD be signed with JWS using the same public key and contain the same set of claims as the access tokens. The authorization server MAY encrypt access tokens and refresh tokens using JWE. Encrypted access tokens MUST be encrypted using the public key of the protected resource. Encrypted refresh tokens MUST be encrypted using the authorization server's public key.

Token introspection allows a protected resource to query the authorization server for metadata about a token. The protected resource makes a request like the following to the token introspection endpoint:

The client assertion parameter is structured as described in . The server responds to an introspection request with a JSON object representing the token containing the following fields as defined in the token introspection specification: Boolean value indicating whether or not this token is currently active at this authorization server. Tokens that have been revoked, have expired, or were not issued by this authorization server are considered non-active. Space-separated list of OAuth 2.0 scope values represented as a single string. Timestamp of when this token expires (integer number of seconds since from 1970-01-01T00:00:00Z UTC) An opaque string that uniquely identifies the user who authorized this token at this authorization server (if applicable) An opaque string that uniquely identifies the OAuth 2.0 client that requested this token The following example is a response from the introspection endpoint:

The authorization server MUST require authentication for both the revocation and introspection endpoints as described in . Protected resources calling the introspection endpoint MUST use credentials distinct from any other OAuth client registered at the server. A protected resource MAY cache the response from the introspection endpoint for a period of time no greater than half the lifetime of the token. A protected resource MUST NOT accept a token that is not active according to the response from the introspection endpoint.

This profile provides RECOMMENDED lifetimes for different types of tokens issued to different types of clients. Specific applications MAY issue tokens with different lifetimes. Any active token MAY be revoked at any time. For clients using the authorization code grant type, access tokens SHOULD have a valid lifetime no greater than one hour, and refresh tokens (if issued) SHOULD have a valid lifetime no greater than twenty-four hours. For clients using the implicit grant type, access tokens SHOULD have a valid lifetime no greater than fifteen minutes. For clients using the client credentials grant type, access tokens SHOULD have a valid lifetime no greater than six hours.

Scopes define individual pieces of authority that can be requested by clients, granted by resource owners, and enforced by protected resources. Specific scope values will be highly dependent on the specific types of resources being protected in a given interface. OpenID Connect, for example, defines scope values to enable access to different attributes of user profiles. Authorization servers SHOULD define and document default scope values that will be used if an authorization request does not specify a requested set of scopes. To facilitate general use across a wide variety of protected resources, authorization servers SHOULD allow for the use of arbitrary scope values at runtime, such as allowing clients or protected resources to use arbitrary scope strings upon registration. Authorization servers MAY restrict certain scopes from use by dynamically registered systems.

A protected resource MUST accept bearer tokens passed in the authorization header as described in . A protected resource MAY also accept bearer tokens passed in the form parameter or query parameter methods. Protected resources MUST define and document which scopes are required for access to the resource.

Protected resources MUST interpret access tokens using either JWT, token introspection, or a combination of the two. The protected resource MUST check the aud (audience) claim, if it exists in the token, to ensure that it includes the protected resource's identifier. The protected resource MUST ensure that the rights associated with the token are sufficient to grant access to the resource. For example, this can be accomplished by querying the scopes associated with the token from the authorization server's token introspection endpoint. A protected resource MUST limit which authorization servers it will accept valid tokens from. A resource server MAY accomplish this using a whitelist of trusted servers, a dynamic policy engine, or other means.

The preceding portions of this OAuth profile provide a level of security adequate for a wide range of use cases, while still maintaining relative ease of implementation and usability for developers, system administrators, and end users. The following are some additional security measures that can be employed for use cases where elevated risks justify the use of additional controls at the expense of implementation effort and usability. This section also addresses future security capabilities, currently in the early draft stages, being added to the OAuth standard suite.

OAuth proof of possession tokens are currently defined in a set of drafts under active development in the Internet Engineering Task Force (IETF) OAuth Working Group. While a bearer token can be used by anyone in possession of the token, a proof of possession token is bound to a particular symmetric or asymmetric key issued to, or already possessed by, the client. The association of the key to the token is also communicated to the protected resource; a variety of mechanisms for doing this are outlined in the draft OAuth 2.0 Proof-of-Possession (PoP) Security Architecture. When the client presents the token to the protected resource, it is also required to demonstrate possession of the corresponding key (e.g., by creating a cryptographic hash or signature of the request). Proof of Possession tokens are somewhat analogous to the Security Assertion Markup Language's (SAML's) Holder-of-Key mechanism for binding assertions to user identities. Proof of possession could prevent a number of attacks on OAuth that entail the interception of access tokens by unauthorized parties. The attacker would need to obtain the legitimate client's cryptographic key along with the access token to gain access to protected resources. Additionally, portions of the HTTP request could be protected by the same signature used in presentation of the token. Proof of possession tokens may not provide all of the same protections as PKI authentication, but they are far less challenging to implement on a distributed scale.

All transactions MUST be protected in transit by TLS as described in BCP195. All clients MUST conform to applicable recommendations found in the Security Considerations sections of and those found in the OAuth 2.0 Threat Model and Security Considerations document.

Transport Layer Security (TLS) and Datagram Transport Layer Security (DTLS) are widely used to protect data exchanged over application protocols such as HTTP, SMTP, IMAP, POP, SIP, and XMPP. Over the last few years, several serious attacks on TLS have emerged, including attacks on its most commonly used cipher suites and their modes of operation. This document provides recommendations for improving the security of deployed services that use TLS and DTLS. The recommendations are applicable to the majority of use cases. Nomura Research Institute, Ltd. Ping Identity Microsoft Google Salesforce Nomura Research Institute, Ltd. Ping Identity Microsoft IllumilaJSON

The OpenID Community would like to thank the following people for their contributions to this specification: Mark Russel, Mary Pulvermacher, David Hill, Dale Moberg, Adrian Gropper, Eve Maler, Danny van Leeuwen, John Moehrke, Aaron Seib, John Bradley, Debbie Bucci, Josh Mandel, and Sarah Squire. The original version of this specification was part of the Secure RESTful Interfaces project from The MITRE Corporation, available online at http://secure-restful-interface-profile.github.io/pages/

Copyright (c) 2017 The OpenID Foundation. The OpenID Foundation (OIDF) grants to any Contributor, developer, implementer, or other interested party a non-exclusive, royalty free, worldwide copyright license to reproduce, prepare derivative works from, distribute, perform and display, this Implementers Draft or Final Specification solely for the purposes of (i) developing specifications, and (ii) implementing Implementers Drafts and Final Specifications based on such documents, provided that attribution be made to the OIDF as the source of the material, but that such attribution does not indicate an endorsement by the OIDF. The technology described in this specification was made available from contributions from various sources, including members of the OpenID Foundation and others. Although the OpenID Foundation has taken steps to help ensure that the technology is available for distribution, it takes no position regarding the validity or scope of any intellectual property or other rights that might be claimed to pertain to the implementation or use of the technology described in this specification or the extent to which any license under such rights might or might not be available; neither does it represent that it has made any independent effort to identify any such rights. The OpenID Foundation and the contributors to this specification make no (and hereby expressly disclaim any) warranties (express, implied, or otherwise), including implied warranties of merchantability, non-infringement, fitness for a particular purpose, or title, related to this specification, and the entire risk as to implementing this specification is assumed by the implementer. The OpenID Intellectual Property Rights policy requires contributors to offer a patent promise not to assert certain patent claims against other contributors and against implementers. The OpenID Foundation invites any interested party to bring to its attention any copyrights, patents, patent applications, or other proprietary rights that may cover technology that may be required to practice this specification.

-2017-05-25 Fixed spec cross-reference -2017-04-18 Changed example to use scopes from HEART OAuth FHIR document -2017-04-10 Require native clients to have unique client IDs (restores previous requirement) Clarified PKCE requirements -2017-02-08 Clarified that RS's need to limit which AS's they take tokens from. Added section and requirements for native applications. -2016-09-19 Reorganized document against different conformance aspects. -2016-08-10 Added PKCE requirement for native application support. Removed client TLS auth. Added reference to JWS/JWE algs registry. Clarified approval UX guidelines. Clarified client types restrictions. -2016-04-30 Added conformance statements. Clarified optionality of "sub" and "aud" claims in JWTs. Add justification for JWT and introspection. Fixed "kid" location information. -2016-02-15 Implementer's Draft 1 -2015-11-30 Clarified client instances. Replaced "mitre.org" with "example.com" (JWTs need to be regenerated). Fixed specification references to new RFCs. Clarified scope flexibility. Clarified dynamic registration requirement. Added some UX requirements and guidance. Added security considerations section and TLS BCP reference. -2015-04-24 Fixed references to make it compile -2015-04-01 Imported content from Secure RESTful OAuth profile.