5. Security Considerations

Since requests to the client registration endpoint result in the transmission of clear-text credentials (in the HTTP request and response), the authorization server MUST require the use of a transport-layer security mechanism when sending requests to the registration endpoint. The server MUST support TLS 1.2 [RFC5246] and MAY support additional transport-layer security mechanisms meeting its security requirements. When using TLS, the client MUST perform a TLS/SSL server certificate check, per RFC 6125 [RFC6125]. Implementation security considerations can be found in Recommendations for Secure Use of TLS and DTLS [BCP195].

For clients that use redirect-based grant types such as "authorization_code" and "implicit", authorization servers MUST require clients to register their redirection URI values. This can help mitigate attacks where rogue actors inject and impersonate a validly registered client and intercept its authorization code or tokens through an invalid redirection URI or open redirector. Additionally, in order to prevent hijacking of the return values of the redirection, registered redirection URI values MUST be one of:

A remote web site protected by TLS (e.g., https://client.example.com/oauth_redirect)
A web site hosted on the local machine using an HTTP URI (e.g., http://localhost:8080/oauth_redirect)
A non-HTTP application-specific URL that is available only to the client application (e.g., exampleapp://oauth_redirect)

Public clients MAY register with an authorization server using this protocol, if the authorization server's policy allows them. Public clients use a "none" value for the "token_endpoint_auth_method" metadata field and are generally used with the "implicit" grant type. Often these clients will be short-lived in-browser applications requesting access to a user's resources and access is tied to a user's active session at the authorization server. Since such clients often do not have long-term storage, it is possible that such clients would need to re-register every time the browser application is loaded. To avoid the resulting proliferation of dead client identifiers, an authorization server MAY decide to expire registrations for existing clients meeting certain criteria after a period of time has elapsed. Alternatively, such clients could be registered on the server where the in-browser application's code is served from, and the client's configuration could be pushed to the browser alongside the code.

Since different OAuth 2.0 grant types have different security and usage properties, an authorization server MAY require separate registrations for a piece of software to support multiple grant types. For instance, an authorization server might require that all clients using the "authorization_code" grant type make use of a client secret for the "token_endpoint_auth_method" but any clients using the "implicit" grant type not use any authentication at the token endpoint. In such a situation, a server MAY disallow clients from registering for both the "authorization_code" and "implicit" grant types simultaneously. Similarly, the "authorization_code" grant type is used to represent access on behalf of an end-user, but the "client_credentials" grant type represents access on behalf of the client itself. For security reasons, an authorization server could require that different scopes be used for these different use cases, and, as a consequence, it MAY disallow these two grant types from being registered together by the same client. In all of these cases, the authorization server would respond with an "invalid_client_metadata" error response.

Unless used as a claim in a software statement, the authorization server MUST treat all client metadata as self-asserted. For instance, a rogue client might use the name and logo of a legitimate client that it is trying to impersonate. Additionally, a rogue client might try to use the software identifier or software version of a legitimate client to attempt to associate itself on the authorization server with instances of the legitimate client. To counteract this, an authorization server MUST take appropriate steps to mitigate this risk by looking at the entire registration request and client configuration. For instance, an authorization server could issue a warning if the domain/site of the logo doesn't match the domain/site of redirection URIs. An authorization server could also refuse registration requests from a known software identifier that is requesting different redirection URIs or a different client URI. An authorization server can also present warning messages to end-users about dynamically registered clients in all cases, especially if such clients have been recently registered or have not been trusted by any users at the authorization server before.

In a situation where the authorization server is supporting open client registration, it must be extremely careful with any URL provided by the client that will be displayed to the user (e.g., "logo_uri", "tos_uri", "client_uri", and "policy_uri"). For instance, a rogue client could specify a registration request with a reference to a drive-by download in the "policy_uri", enticing the user to click on it during the authorization. The authorization server SHOULD check to see if the "logo_uri", "tos_uri", "client_uri", and "policy_uri" have the same host and scheme as the those defined in the array of "redirect_uris" and that all of these URIs resolve to valid web pages. Since these URI values that are intended to be displayed to the user at the authorization page, the authorization server SHOULD protect the user from malicious content hosted at the URLs where possible. For instance, before presenting the URLs to the user at the authorization page, the authorization server could download the content hosted at the URLs, check the content against a malware scanner and blacklist filter, determine whether or not there is mixed secure and non-secure content at the URL, and other possible server-side mitigations. Note that the content in these URLs can change at any time and the authorization server cannot provide complete confidence in the safety of the URLs, but these practices could help. To further mitigate this kind of threat, the authorization server can also warn the user that the URL links have been provided by a third party, should be treated with caution, and are not hosted by the authorization server itself. For instance, instead of providing the links directly in an HTML anchor, the authorization server can direct the user to an interstitial warning page before allowing the user to continue to the target URL.

Clients MAY use both the direct JSON object and the JWT-encoded software statement to present client metadata to the authorization server as part of the registration request. A software statement is cryptographically protected and represents claims made by the issuer of the statement, while the JSON object represents the self-asserted claims made by the client or developer directly. If the software statement is valid and signed by an acceptable authority (such as the software API publisher), the values of client metadata within the software statement MUST take precedence over those metadata values presented in the plain JSON object, which could have been intercepted and modified.

Like all metadata values, the software statement is an item that is self-asserted by the client, even though its contents have been digitally signed or MACed by the issuer of the software statement. As such, presentation of the software statement is not sufficient in most cases to fully identify a piece of client software. An initial access token, in contrast, does not necessarily contain information about a particular piece of client software but instead represents authorization to use the registration endpoint. An authorization server MUST consider the full registration request, including the software statement, initial access token, and JSON client metadata values, when deciding whether to honor a given registration request.

If an authorization server receives a registration request for a client that is not intended to have multiple instances registered simultaneously and the authorization server can infer a duplication of registration (e.g., it uses the same "software_id" and "software_version" values as another existing client), the server SHOULD treat the new registration as being suspect and reject the registration. It is possible that the new client is trying to impersonate the existing client in order to trick users into authorizing it, or that the original registration is no longer valid. The details of managing this situation are specific to the authorization server deployment and outside the scope of this specification.

Since a client identifier is a public value that can be used to impersonate a client at the authorization endpoint, an authorization server that decides to issue the same client identifier to multiple instances of a registered client needs to be very particular about the circumstances under which this occurs. For instance, the authorization server can limit a given client identifier to clients using the same redirect-based flow and the same redirection URIs. An authorization server SHOULD NOT issue the same client secret to multiple instances of a registered client, even if they are issued the same client identifier, or else the client secret could be leaked, allowing malicious impostors to impersonate a confidential client.