A set of SASL and GSS-API Mechanisms for OAuthYahoo! Inc.wmills@yahoo-inc.com tjs@psaux.comNokia Siemens NetworksLinnoitustie 6Espoo02600Finland+358 (50) 4871445Hannes.Tschofenig@gmx.nethttp://www.tschofenig.priv.atKITTEN
OAuth enables a third-party
application to obtain limited access to a protected resource, either on
behalf of a resource owner by orchestrating an approval interaction, or by allowing the
third-party application to obtain access on its own behalf.
This document defines how an application client uses credentials obtained via OAuth
over the Simple Authentication and Security Layer (SASL) or the
Generic Security Service Application Program Interface (GSS-API)
to access a protected resource at a resource serve. Thereby, it enables
schemes defined within the OAuth framework for non-HTTP-based application protocols.
Clients typically store the user's long term credential. This does, however, lead to
significant security vulnerabilities, for example, when such a
credential leaks. A significant benefit of OAuth for usage in
those clients is that the password is replaced by a
token. Tokens typically provided limited access rights and can
be managed and revoked separately from the user's long-term credential (password).
OAuth enables a third-party
application to obtain limited access to a protected resource, either on
behalf of a resource owner by orchestrating an approval interaction, or by allowing the
third-party application to obtain access on its own behalf. The core OAuth
specification does not define the interaction between the
client and the resource server with the access to a protected resource using an Access Token.
This functionality is described in separate specifications, for example Bearer tokens , MAC tokens ,
and OAuth 1.0a . In each of these are defined in an HTTP-based
environment only.
shows the abstract message flow as shown in Figure 1 of .This document takes
advantage of the OAuth protocol and its deployment base to provide a way to use
SASL as well as the GSS-API to gain
access to resources when using non-HTTP-based protocols, such as the Internet Message
Access Protocol (IMAP) and SMTP ,
which is what this memo uses in the examples.The Simple Authentication and Security Layer (SASL) is a framework
for providing authentication and data security services in
connection-oriented protocols via replaceable mechanisms. It
provides a structured interface between protocols and mechanisms.
The resulting framework allows new protocols to reuse existing
mechanisms and allows old protocols to make use of new mechanisms.
The framework also provides a protocol for securing subsequent
protocol exchanges within a data security layer.
The Generic Security Service Application Program Interface (GSS-API)
provides a framework for applications to support multiple
authentication mechanisms through a unified interface.
This document
defines SASL mechanisms for OAuth, and it conforms to the new
bridge between SASL and the GSS-API called GS2 . This means
that this document defines both SASL and GSS-API
mechanisms. Implementers may be interested in
either the SASL, the GSS-API, or even both mechanisms. To
faciliate these two variants, the description has been split into two
parts, one part that provides normative references for those interested in
the SASL OAuth mechanism (see ), and a second part
for those implementers that wish to implement the GSS-API portion (see
).
When OAuth is integrated into SASL and the GSS-API the high-level steps are as follows:
(A) The client requests authorization from the resource owner. The
authorization request can be made directly to the resource owner
(as shown), or preferably indirectly via the authorization
server as an intermediary. (B) The client receives an authorization grant which is a credential
representing the resource owner's authorization, expressed using
one of four grant types defined in this specification or using
an extension grant type. The authorization grant type depends
on the method used by the client to request authorization and
the types supported by the authorization server. (C) The client requests an access token by authenticating with the
authorization server and presenting the authorization grant. (D) The authorization server authenticates the client and validates
the authorization grant, and if valid issues an access token. (E) The client requests the protected resource from the resource
server and authenticates by presenting the access token. (F) The resource server validates the access token, and if valid,
indicates a successful authentication.Steps (E) and (F) are not defined in and are the
main functionality specified within this document.
Consequently, the message exchange shown in is the result of this specification.
The client will genrally need to determine the authentication endpoints (and
perhaps the service endpoints) before the OAuth 2.0 protocol
exchange messages in steps (A)-(D) are executed. The discovery of the resource owner
and authorization server endpoints is
outside the scope of this specification. The client must discover those
endpoints using a discovery mechanisms such as Webfinger using host-meta . In band discovery is not tenable if
clients support the OAuth 2.0 password grant. Once credentials are obtained the
client proceeds to steps (E) and (F) defined in this specification.
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
.The reader is assumed to be familiar with the terms used in the OAuth 2.0 specification .In examples, "C:" and "S:" indicate lines sent by the client and server respectively. Line
breaks have been inserted for readability.Note that the IMAP SASL specification requires base64 encoding message, not this memo.SASL is used as a generalized authentication method in a variety of application layer protocols. This
document defines the following SASL mechanisms for usage with OAuth:
Authorization using Bearer tokens.
Authorization using OAuth 1.0a tokens.
Adds channel binding capability
to OAUTH10A for additional security guarantees.
Any new OAuth token scheme MAY define a new SASL mechanism compatible with the
mechanisms defined here by simply registering the new name(s) and citing this
specification for the further definition. New channel binding enabled "-PLUS"
mechanisms defined in this way MUST include message integrity protection. A
newly defined mechanism would also need to register a new GS2 OID.
These mechanisms are client initiated and lock-step, the server always replying to a client
message. In the case where the client has and correctly uses a valid token the flow is:
Client sends a valid and correct initial client response.
Server responds with a successful authentication.
In the case where authorization fails the server sends an error result, then client MUST
then send an additional message to the server in order to allow the server to finish the
exchange. Some protocols and common SASL implementations do not support both sending a SASL
message and finalizing a SASL negotiation, the additional client message in the error case
deals with this problem. This exchange is:
Client sends an invalid initial client response.Server responds with an error message. Client sends a dummy client reponse.Server fails the authentication.Client responses are a key/value pair sequence. The initial client response
includes a gs2-header as defined in GS2 , which carries the
authorization ID. These key/value pairs carry the equivalent values from
an HTTP context in order to be able to complete an OAuth style HTTP authorization.
The ABNF syntax is:
The following key/value pairs are defined in the client response:
The payload of the HTTP Authorization header
for an equivalent HTTP OAuth authroization.Contains the host name to which the client connected.Contains the port number represented as a
decimal positive integer string without leading zeros
to which the client connected.The HTTP query string. In non-channel binding mechanisms
this is reserved, the client SHOUD NOT send it, and has the default value
of "". In "-PLUS" variants this carries
a single key value pair "cbdata" for the channel binding data payload formatted
as an HTTP query string.
In authorization schemes that use signatures, the client MUST send host and
port number key/values, and the server MUST fail an authorization request requiring
signatures that does not have host and port values. For authorization schemes
that require a URI scheme as part of the data being signed "http" is always used. In
OAuth 1.0a for example, the signature base string includes the reconstructed HTTP
URL.
In these mechanisms values for path, query string and post body are
assigned default values. OAuth authorization schemes MAY define usage of
these in the SASL context and extend this specification. For OAuth
schemes that use request signatures the default values MUST be used unless
explict values are provided in the client response. The following key
values are reserved for future use:
HTTP method for use in signatures, the default value is "POST".
HTTP path data, the default value is "/".
HTTP post data, the default value is "".
The OAuth scheme related mechanisms are also GSS-API mechanisms, see
for further detail.
The gs2-header is used as follows:
The "gs2-nonstd-flag" MUST NOT be present.
The "gs2-authzid" carries the authorization identity as
specified in [RFC5801]. If present the application MUST determine
whether access is granted for the identity asserted in the
OAuth credential, if it does not the server
MUST fail the negotiation.
In the non "-PLUS" mechanisms the "gs2-cb-flag" MUST be set to "n" because
channel-binding [RFC5056] data is not expected.
In the OAUTH10A-PLUS mechanism (or other -PLUS variants based on this specification)
the "gs2-cb-flag" MUST be set appropriately by the client.
The server validates the response per the specification for the authorization
scheme used. If the authorization scheme used includes signing of the request
parameters the client must provide a client response that satisfies
the data requirements for the scheme in use.
In a "-PLUS" mechanism the server examines the channel binding data,
extracts the channel binding unique prefix, and extracts the raw channel biding
data based on the channel binding type used. It then computes it's own copy of
the channel binding payload and compares that to the payload sent by the client in
the cbdata key/value. Those two must be equal for channel binding to succeed.
The server responds to a successfully verified client message by completing the SASL
negotiation. The authenticated identity reported by the SASL mechanism is the
identity securely established for the client with
the OAuth credential. The application, not the SASL mechanism, based on local
access policy determines whether the identity reported by the mechanism
is allowed access to the requested resource. Note that the semantics of the
authz-id is specified by the SASL framework .
Some OAuth schemes can carry both an owner or resource identity and a "proxy" identity,
for example an OAuth 1.0a
mechanism where the consumer key (oauth_consumer_key) identifies the entity
using the token and the token itself identifies the owner or resouce. If both identities
are needed by an application the developer will need to provide a way to
communicate that from the SASL mechanism back to the application such as
a GSS-API named type like GSS_C_NT_USER_NAME or
a comparable newly defined GSS-API name type or name attribute .
The identity asserted by the OAuth authorization server is canonical for display.
The server MAY provide a different canonical form based on local data.
For a failed authentication the server returns a JSON
formatted error result, and fails the authentication. The error result consists
of the following values:
The authorization error code. Valid error codes are
defined in the IANA [[need registry name]] registry
specified in the OAuth 2 core specification.
An OAuth scope which is valid to access the service.
This may be empty which implies that unscoped tokens are required,
or a space separated list. Use of a space separated list is
NOT RECOMMENDED.
If the resource server provides a scope the client SHOULD always request scoped
tokens from the token endpoint. The client MAY use a scope other than the one
provided by the resource server. Scopes other than those advertised by the
resource server are be defined by the resource owner and provided in service
documentation or discovery information (which is beyond the scope of this memo).
If not present then the client SHOULD presume an empty scope (unscoped token) is needed.
If channel binding is in use and the channel
binding fails the server responds with a status code set to 412 to indicate that the channel
binding precondition failed. If the authentication scheme in use does not include
signing the server SHOULD revoke the presented credential and the client SHOULD
discard that credential.
Section 3.6 of explicitly prohibits additional information
in an unsuccessful authentication outcome. Therefor, the error
message is sent in a normal message. The client MUST then send an
additional client response consisting of a single %x01 (control A) character to
the server in order to allow the server to finish the exchange.
Some OAuth mechanisms support authorization using signatures, which requires that both client and
server construct the string to be signed. OAuth 2 is designed for
authentication/authorization to access specific URIs. SASL is designed for user authentication,
and has no facility for being more specific. In this mechanism we require or
define default values for the data elements from an HTTP request which allow the
signature base string to be constructed properly.
The default HTTP path is "/" and the default post body is empty. These atoms are
defined as extension points so
that no changes are needed if there is a revision of SASL which supports more
specific resource authorization, e.g. IMAP access to a specific folder or FTP access
limited to a specific directory. Using the example in the OAuth 1.0a specification
as a starting point, on an IMAP server running on port 143 and given
the OAuth 1.0a style authorization request (with %x01 shown as ^A and line breaks added
for readability) below:
The signature base string would be constructed per the OAuth 1.0
specification with the following things noted:
The method value is defaulted to POST.The scheme defaults to be "http", and any port number other than 80 is included.The path defaults to "/".The query string defaults to "".
In this example the signature base string with line breaks added for
readability would be:
The channel binding data is carried in the "qs" (query string) key value pair
formatted as a standard HTTP query parameter with the name "cbdata". Channel
binding requires that the channel binding data be integrity protected
end-to-end in order to protect against man-in-the-middle attacks. All authorization
schemes offered with "-PLUS" mechanisms MUST provide integrity protection. It
should be noted that while the Bearer token scheme specifies SSL for normal usage it
offers no integrity protection and is not suitable for use with channel binding.The channel binding data is computed by the client based on it's choice of
preferred channel binding type. As specified in , the
channel binding information MUST start with the channel binding unique prefix, followed
by a colon (ASCII 0x3A), followed by a base64 encoded channel binding
payload. The channel binding payload is the raw data from the channel binding
type. For example, if the client is using tls-unique for channel binding then
the raw channel binding data is the TLS finished message as specified in section 3.1 of
. Note: The normative references in this section are informational for SASL
implementers, but they are normative for GSS-API implementers.A SASL OAuth mechanism is also a GSS-API mechanism and the messages
described in are the same with the following
changes to the GS2 related elements:
the GS2 header on the client's first message is excluded when used as
a GSS-API mechanism.
the initial context token
header is prefixed to the client's first authentication message
(context token), as described in Section 3.1 of RFC 2743,
The GSS-API mechanism OIDs are:
OAUTHBEARER: [[TBD: IANA -- probably in the 1.3.6.1.5.5 tree]]OAUTH10A: [[TBD: IANA -- probably in the 1.3.6.1.5.5 tree]]OAuth mechanims security contexts always have the mutual_state flag
(GSS_C_MUTUAL_FLAG) set to TRUE. OAuth supports credential
delegation, therefore security contexts may have the
deleg_state flag (GSS_C_DELEG_FLAG) set to either TRUE or FALSE.
The mutual authentication property of this mechanism relies on
successfully comparing the TLS server identity with the negotiated
target name. Since the TLS channel is managed by the application
outside of the GSS-API mechanism, the mechanism itself is unable to
confirm the name while the application is able to perform this
comparison for the mechanism. For this reason, applications MUST
match the TLS server identity with the target name, as discussed in
.OAuth mechanisms do not support per-message tokens or
GSS_Pseudo_random.OAuth supports a standard generic name syntax for acceptors, such as
GSS_C_NT_HOSTBASED_SERVICE (see , Section 4.1).
These service names MUST be associated with the "entityID" claimed by
the RP.
OAuth mechanisms support only a single name type for initiators:
GSS_C_NT_USER_NAME. GSS_C_NT_USER_NAME is the default name type.
The query, display, and exported name syntaxes for OAuth principal
names are all the same. There is no OAuth-specific name syntax;
applications SHOULD use generic GSS-API name types, such as
GSS_C_NT_USER_NAME and GSS_C_NT_HOSTBASED_SERVICE (see ,
Section 4). The exported name token does, of course, conform to
, Section 3.2, but the "NAME" part of the token should be
treated as a potential input string to the OAuth name normalization
rules.
These examples illustrate exchanges between an IMAP and SMTP clients and servers.Note to implementers: Authorization scheme names are case insensitive. One example
uses "Bearer" but that could as easily be "bearer", "BEARER", or "BeArEr".
This example shows a successful OAuth 2.0 bearer token exchange. Note that line
breaks are inserted for readability.As required by IMAP , the payloads are base64-encoded. The
decoded initial client response (with %x01 represented as ^A and long lines
wrapped for readability) is:
The same credential used in an SMTP exchange is shown below.
Note that line breaks are inserted for readability, and that the
SMTP protocol terminates lines with CR and LF characters (ASCII values
0x0D and 0x0A), these are not displayed explicitly in the example.This example shows channel binding in the context of an OAuth 1.0a signed
authorization request. Note that line breaks are inserted for
readability.As required by IMAP , the payloads are
base64-encoded. The
decoded initial client response (with %x01 represented as ^A and lines
wrapped for readability) is:
In this example the signature base string with line breaks added for
readability would be:
This example shows a failed exchange because of the empty Authorization header, which is
how a client can query for the needed scope. Note that line breaks are inserted for
readability. The decoded initial client response is: The decoded server error response is: The client responds with the required dummy response.
This example shows a channel binding failure in an empty request.
The channel binding information is empty. Note that line breaks are inserted for
readability. The decoded initial client response is: The decoded server response is: The client responds with the required dummy response.
This example shows an authorization failure in an SMTP exchange.
Note that line breaks are inserted for readability, and that the
SMTP protocol terminates lines with CR and LF characters (ASCII values
0x0D and 0x0A), these are not displayed explicitly in the example.The server returned an error message in the 334 SASL message, the
client responds with the required dummy response, and
the server finalizes the negotiation.
This mechanism does not provide a security layer, but does provide a provision for
channel binding. The OAuth 2 specification allows for a variety of usages, and the security properties
of these profiles vary. The usage of bearer tokens, for example, provide security features
similar to cookies. Applications using this mechanism SHOULD exercise the same level of care
using this mechanism as they would in using the SASL PLAIN mechanism. In
particular, TLS 1.2 or an equivalent secure channel MUST be implemented and its
usage is RECOMMENDED.
The channel binding in this mechanism has different properties based on the authentication
scheme used. The integrity guarantee for channel binding depends on the quality of the
guarantee in the the authorization scheme.
It is possible that SASL will be authenticating a connection and the life of that
connection may outlast the life of the token used to authenticate
it. This is a common problem in application protocols where connections are long-lived, and
not a problem with this mechanism per se. Servers MAY unilaterally disconnect
clients in accordance with the application protocol.
An OAuth credential is not equivalent to the password or primary account
credential. There are protocols like XMPP that allow actions like change
password. The server SHOULD ensure that actions taken in the authenticated
channel are appropriate to the strength of the presented credential.
Tokens have a lifetime associated with them. Reducing the lifetime of a
token provides security benefits in the case that tokens leak. In addition a previously obtained
token MAY be revoked or rendered invalid at any time. The client MAY request a new access token for each
connection to a resource server, but it SHOULD cache and re-use access credentials that appear
to be valid. The IANA is requested to register the following SASL profile: SASL mechanism profile: OAUTHBEARERSecurity Considerations: See this documentPublished Specification: See this documentFor further information: Contact the authors of this document.Owner/Change controller: the IETFNote: None The IANA is requested to register the following SASL profile: SASL mechanism profile: OAUTH10ASecurity Considerations: See this documentPublished Specification: See this documentFor further information: Contact the authors of this document.Owner/Change controller: the IETFNote: None The IANA is requested to register the following SASL profile: SASL mechanism profile: OAUTH10A-PLUSSecurity Considerations: See this documentPublished Specification: See this documentFor further information: Contact the authors of this document.Owner/Change controller: the IETFNote: NoneIANA is further requested to assign an OID for thESE GSS mechanismS
in the SMI numbers registry, with the prefix of
iso.org.dod.internet.security.mechanisms (1.3.6.1.5.5) and to
reference this specification in the registry.
The authors would like to thank the members of the Kitten working group, and in
addition and specifically: Simon Josefson, Torsten Lodderstadt, Ryan Troll, and Nico Williams.
[[ to be removed by RFC editor before publication as an RFC ]]
-08
Fixed the channel binding examples for p=$cbtype
More tuning of the authcid language and edited and renamed 3.2.1.
-07
Struck the MUST langiage from authzid.
-06
Removed the user field. Fixed the examples again.
Added canonicalization language.
-05
Fixed the GS2 header language again.
Separated out different OAuth schemes into different SASL mechanisms. Took out the
scheme in the error return. Tuned up the IANA registrations.
Added the user field back into the SASL message.
Fixed the examples (again).
-04
Changed user field to be carried in the gs2-header, and made gs2 header explicit in all cases.
Converted MAC examples to OAuth 1.0a. Moved MAC to an informative reference.
Changed to sending an empty client response (single control-A) as the second message of a failed sequence.
Fixed channel binding prose to refer to the normative specs and removed the hashing of large channel
binding data, which brought mroe problems than it solved.
Added a SMTP examples for Bearer use case.
-03
Added user field into examples and fixed egregious errors there as well.
Added text reminding developers that Authorization scheme names are case insensitive.
-02
Added the user data element back in.
Minor editorial changes.
-01
Ripping out discovery. Changed to refer to I-D.jones-appsawg-webfinger instead
of WF and SWD older drafts.
Replacing HTTP as the message format and adjusted all examples.
-00
Renamed draft into proper IETF naming format now that it's adopted.
Minor fixes.