| TRAM Working Group | A. Johnston |
| Internet-Draft | Avaya |
| Intended status: Standards Track | J. Uberti |
| Expires: February 21, 2015 | |
| J. Yoakum | |
| K. Singh | |
| Avaya | |
| August 20, 2014 |
An Origin Attribute for the STUN Protocol
draft-ietf-tram-stun-origin-01
STUN, or Session Traversal Utilities for NAT, is a protocol used to assist other protocols traverse Network Address Translators or NATs. STUN, and STUN extensions such as TURN, or Traversal Using Relays around NAT, and ICE, Interactive Communications Establishment, have been around for many years but with WebRTC, Web Real-Time Communications, STUN and related extensions are about to see major deployments and implementation due to these protocols being implemented in browsers. This specification defines an ORIGIN attribute for STUN that can be used in similar ways to the HTTP header field of the same name. WebRTC browsers utilizing STUN and TURN would include this attribute which would provide servers with additional information about the STUN and TURN requests they receive. This specification defines the usage of the STUN ORIGIN attribute for web and SIP contexts.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on February 21, 2015.
Copyright (c) 2014 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
STUN, or Session Traversal Utilities for NAT, is a protocol used to assist other protocols traverse Network Address Translators or NATs. TURN, or Traversal Using Relays around NAT [RFC5766], is a STUN extension [RFC5389] that allows endpoints to acquire a relayed address for media flows. It is most commonly used in conjunction with ICE, Interactive Connectivity Establishment [RFC5245], which is used to establish peer-to-peer flows between endpoints through NATs and firewalls.
STUN defines three authentication modes, depending on the STUN usage. For STUN binding requests sent between peers, such as for ICE connectivity checks, a short term authentication method is recommended. Each peer contributes random strings which are exchanged over signaling and used to authenticate the connectivity checks. For TURN, a usage of STUN used to acquire and refresh relay addresses, a long term authentication method is recommended. This authentication is similar to SIP Digest [RFC3261], which involves an authentication challenge for each request. A server, upon receipt of a TURN request, generates an authentication challenge that includes a realm and nonce. The client resends the TURN request supplying a user name and password based on the realm indicated by the server. For a STUN binding request sent to a STUN server, no authentication is recommended, as generating the response is less work for a server than the server utilizing the short term or long term authentication approach. In addition, the resource requirements of operating a STUN server are minimal.
WebRTC, Web Real-Time Communications, adds peer-to-peer real-time, interactive voice and video media capabilities and data channels to browsers [I-D.ietf-rtcweb-overview] without a plugin or download, and allows web developers to access this functionality using JavaScript API calls [WebRTC-API]. WebRTC includes STUN, TURN, and ICE client functionality built into browsers. For a session established between two browsers, if either browser is behind a NAT, a STUN server is necessary. Public STUN servers are currently available and a web application can suggest a particular STUN server be used. In other cases, a TURN server is needed to establish a peer connection. In this case, TURN credentials need to be available to the browser for the long term authentication approach. A TURN server for WebRTC might serve a number of different domains and realms.
From the perspective of the web application provider, providing service for a number of different domains and realms, it is useful to know something about the source of the STUN request when processing the request. For a web application provider STUN or TURN server, the server will have no idea which web pages or sites are sending binding requests to the service. In conventional applications, the SOFTWARE attribute would provide some identifying information to the service, but that no longer works when the browser is the application. For a web application provider TURN server, the TURN server does not know which realm to include in an authentication challenge.
In the web world, HTTP requests have the concept of origin. The origin of a web page, as defined in [RFC6454], is defined by the URI's scheme, host or IP address, and port portions. The HTTP Origin header field inserted by the web browser carries this information and is useful information for servers that receive HTTP requests generated via JavaScript. For example, Cross Origin Resource Sharing, CORS, allows an HTTP server to serve HTTP requests from multiple origins.
This specification proposes extending the origin concept to STUN requests. STUN requests generated by a web browser would include the origin of the HTTP page that is initiating the Peer Connection. Using this extra information, a STUN server could use the origin to determine which STUN binding requests to respond to, reducing the load on a STUN server. Using this information, a TURN server could use the origin to determine which realm to include in the authentication challenge. A TURN server can also use the origin information for logging and analytics, and also as additional information after authentication for providing service.
An important use case that the STUN Origin helps solve is the operation of a multi-tenanted TURN server (i.e. a TURN server that serves multiple, perhaps tens of thousands of different domains). The problem associated with this use case is described in Section 4.5 of [I-D.ietf-tram-auth-problems]. While it is possible for a TURN server to use the same authentication credentials across many domains, a more likely (and more manageable) scenario is to have separate credentials for each domain, and hence a different realm for each domain. To implement this, a TURN server needs to know which realm to include in authentication challenge to TURN clients. One way to do this would be to create a unique TURN URI for each realm. This would require either a separate IP address or port for each realm, and this unique URI would need to be correctly provisioned by each domain (i.e. included in JavaScript, which then could not be copied between domains). Clearly, this doesn't scale for hundreds or thousands of domains. Origin information solves this problem since TURN requests will contain the domain in the Origin attribute. The TURN server just needs to be configured with a mapping between a domain (conveyed in the Origin) and the realm string (to be used in the authentication challenge). Thus, a single TURN URI could be used across all domains, and the resulting JavaScript code would be portable. There is no need for thousands of IP addresses or ports to be allocated and managed.
It has been suggested that this origin insight is not needed if the server_name TLS extension in [RFC6066] is supported. This extension allows a TLS client to provide to the TLS server the name of the server they are contacting. In this case, the STUN or TURN client using TLS transport would provide the domain from the TURN server URI during the TLS client hello, allowing the TURN server to respond with the appropriate server certificate. For the TURN server domain to be used by the TURN server to choose the appropriate realm, this would require a unique TURN URI to be provisioned per realm. This is not scalable for supporting thousands of realms. Also, this URI would need to be provisioned in the JavaScript, making the resulting code non-portable. Finally, [RFC6066] provides no help for UDP or TCP transport, which are the most commonly used transports today for STUN and TURN.
Another approach that could be pursued is for the client to be explicitly provisioned with a realm value, to which its username and password are scoped. When using the long-term authentication method to authenticate to a TURN server, the client would include this REALM value in the initial, unauthorized requests, allowing the TURN server to know which REALM to use in its authorization challenge. This approach avoids many of the issues with the RFC 6066 approach, but it still requires the realm value to be explicitly provisioned in Javascript. In addition, it does not work in unauthenticated usages, i.e. STUN binding requests sent to a STUN server.
Note that the origin information is most useful as a hint in initial STUN and TURN requests as received by a server. However, origin information still has value throughout the session even after authentication for logging and other purposes.
The following sections of this document define the STUN ORIGIN attribute and define its usage.
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 RFC 2119 [RFC2119].
This specification defines how to apply the web origin concept and syntax of [RFC6454] to the STUN protocol.
This specification defines a new Attribute to the STUN protocol [RFC5389]. The attribute is called ORIGIN and uses the syntax defined in Section 15 of [RFC5389]. The number used for this in the type field is 0x802F, chosen in the comprehension optional range. The value of ORIGIN is a variable-length value. It MUST contain a UTF-8 [RFC3629] encoded sequence of characters. Senders MAY include multiple ORIGIN attributes in a request, and receivers MUST support parsing and receiving multiple ORIGIN attributes. The size and number of ORIGINs included in a STUN message can have a major impact on the size of a STUN packet, and could potentially cause UDP fragmentation. HTTP origins are less than 267 characters (the maximum 253 character domain name plus 8 characters for the URI scheme plus 5 characters for the port number). It is NOT RECOMMENDED that an ORIGIN attribute be included if the total size (including type, length, and padding) is larger than 272 octets. Multiple ORIGIN attributes SHOULD NOT be included if the total size of these attributes exceeds 272 octets.
Editor's Note: At the appropriate time, the authors will work with the chairs of TRAM to follow [RFC4020] procedures to ensure that no attribute collisions occur while running code is being developed and tested.
For a web browser (HTTP User Agent), the contents of the ORIGIN attribute is the unicode-serialization of an origin defined in Section 6.1 of [RFC6454]. The origin value included is the same as the Origin header field for an HTTP request generated from the web page that is creating the Peer Connection. It does not include any string terminating (\x00) character in the serialization.
For a SIP User Agent [RFC3261] using STUN and TURN, the ORIGIN attribute is set to be the URI of the registrar server used by the User Agent (i.e. the Request-URI of a REGISTER method).
For an XMPP client [RFC6120] using STUN and TURN, the ORIGIN attribute is an XMPP URI [RFC5122] representing the domainpart of the client's Jabber ID (JID) [RFC6122]; for example, if the client's JID is "juliet@im.example.com/balcony" then the ORIGIN attribute would be "xmpp:im.example.com".
Other contexts can define a usage of the ORIGIN attribute to use an appropriate URI or URL.
To ensure backwards compatibility with [RFC3489], the ORIGIN attribute is padded to ensure its length is a multiple of 4 octets.
If an ORIGIN attribute is not present in a request, it is up to the server how to handle the request. For example, it could assume a default Origin.
For STUN requests sent without authentication to a STUN server (i.e. STUN binding requests sent to a STUN server), the STUN client SHOULD include the ORIGIN attribute. A STUN server can derive additional information for logging and analytics about the request through the ORIGIN attribute, such as the source of the request. For example, an enterprise STUN server might only reply to STUN binding requests from certain domains.
For STUN requests sent using the long-term authentication method, such as TURN [RFC5766] allocate requests, the STUN client SHOULD include the ORIGIN attribute. Including the ORIGIN attribute in the Send method is NOT RECOMMENDED. A TURN server can use the ORIGIN attribute to determine which REALM to include in the authentication challenge. A TURN server can also use the ORIGIN attribute after authentication to provide appropriate service. See the section below on "Multiple Origins."
For the NAT Behavior Discovery Usage in [RFC5780], the ORIGIN attribute SHOULD be included in requests sent to a STUN server. This usage is most similar to the STUN Usage described earlier.
For STUN requests sent using the short-term authentication method, such as ICE connectivity checks [RFC5245], the use of the ORIGIN attribute is NOT RECOMMENDED. No valid use cases for the ORIGIN attribute have been identified to date.
For media keep-alive STUN requests described in Section 20 of [RFC5245], the use of the ORIGIN attribute is NOT RECOMMENDED. No valid use cases for the ORIGIN attribute have been identified to date.
For SIP keep-alive STUN requests described in [RFC5626], the use of the ORIGIN attribute is NOT RECOMMENDED. No valid use cases for the ORIGIN attribute have been identified to date.
Multiple Origins for HTTP Requests are described in Section 7.2 of [RFC6454]. As a result, a browser MAY generate a STUN request with multiple ORIGIN attributes present. A browser MUST NOT include multiple origins in a single ORIGIN attribute and instead use one Origin per attribute. If the ORIGIN attribute is being used by a TURN server to select the realm for authentication, the TURN server MAY use any or all of the Origins to select the realm. Any future new usage scenarios of the ORIGIN attribute need to describe how to handle multiple Origins.
This specification, if approved, adds a new value to the IANA "STUN Attributes Registry" created by [RFC5389]. The ORIGIN attribute value is 0x802F.
The security considerations of [RFC6454] apply to this extension. Servers using the information present in the STUN ORIGIN attribute need to realize that this attribute could be set arbitrarily by a non-browser client or modified by an intermediary. The method proposed in this document is not meant to replace existing STUN authentication mechanisms but to provide additional information to the server for logging and analytics and how to handle the request after authentication.
Just as browsers do not allow a web application to set the Origin header field via JavaScript, browsers should not allow a web application through JavaScript to set the STUN ORIGIN attribute.
While the STUN MESSAGE-INTEGRITY attribute can provide integrity protection for all attributes present in a STUN request, MESSAGE-INTEGRITY is not present in the initial STUN message sent. As a result, an ORIGIN attribute could be modified or removed from a STUN request without the server knowing. DTLS or TLS transport SHOULD be used when integrity protection for the ORIGIN attribute is important.
The STUN ORIGIN attribute does have privacy implications. The recipient of the STUN request learns the web origin of the user. In addition, an on-path attacker could determine this information by inspecting STUN messages between the STUN client and STUN server, depending on the transport used. This information is often available in other messages sent by the browser, such as DNS or HTTP requests. However, in cases where secure HTTP is used, including the ORIGIN attribute over an unencrypted transport could leak this information. STUN has a defined TLS transport; however, TLS transport is generally unsuitable for the real-time media flows that follow STUN requests and must use the same transport. The DTLS transport for STUN [I-D.ietf-tram-stun-dtls] provides a very good privacy solution to this problem. In cases where privacy is paramount, the ORIGIN attribute SHOULD NOT be included or only included if DTLS or TLS transport is used.
Note to RFC Editor: Please remove this entire section prior to publication, including the reference to RFC 6982.
This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in [RFC6982]. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.
According to [RFC6982], "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".
Two proof-of-concept implementations have been created in support of this proposed standard. One provides a WebRTC enabled browser that includes the appropriate STUN ORIGIN Attribute with the Origin insight known to the browser in STUN/TURN messages sent to servers. The other provides an example of a multiple realms capable TURN server that takes advantage of Origin insight provided in the STUN ORIGIN Attribute.
A Chrome browser implementation has been created by Graham Yoakum and Ryan Yoakum (Skobalt LLC) and is freely licensed under the standard terms of the open source Chromium and WebRTC projects. This proof-of-concept version of the Google Chrome browser (nicknamed 'Chromeo') sends Origin insight in STUN and TURN messages using the proposed new STUN ORIGIN attribute with a value of 0x802F (as initially proposed, however that value is easily changed in a single line of code). 'Chromeo' includes a Chrome flag to enable and disable this unique feature (and is by default disabled to prevent any non-intentional use of this feature until the standard is finalized). This implementation is based on is draft-johnston-tram-stun-origin-02.
Coordinated changes to both the WebRTC and Chromium open source projects have been formally submitted for consideration. The two submitted change lists together implement the complete browser proof-of-concept. 'Chromeo' has been built for Linux and STUN protocol behavior has been verified using WireShark traces illustrating that proper STUN Origin attributes are being included in STUN/TURN messages sent by the browser to servers (screen captures of STUN messages illustrating the Origin attribute and content are available).
The WebRTC and Chromium open source projects can be found at: http://www.webrtc.org/ and http://www.chromium.org/
Google can choose to accept or modify the changes proposed for Chrome and other browser vendors can access and take advantage of the publicly available WebRTC and Chromium open source submissions as desired. Hopefully this will enable browsers to quickly implement STUN Origin enhancements.
A multiple realms capable advanced open source Origin enabled TURN server (named 'Coturn') has been created by Oleg Moskalenko and is freely licensed under the New BSD license. This reference implementation and proof-of-concept provides a clone (a spin-off) of the rfc5766-turn-server project adding Origin-based multiple realms support.
'Coturn' is backward-compatible with rfc5766-turn-server project but the code is more complex and it uses a different (also more complex) database structure. It is the intent to add all IETF TRAM TURN server related capabilities to this project as they mature. 'Coturn' is publicly available and can be found at: https://code.google.com/p/coturn/
Thanks to John Selbie, Tirumaleswar Reddy, Simon Perreault, Marc Petit-Huguenin, Andy Hutton, and Oleg Moskalenko for their feedback and reviews. Special thanks to Graham Yoakum and Ryan Yoakum of Skobalt LLC and Oleg Moskalenko of rfc5766-turn-server project for contributing open source proof-of-concept implementations for a Chrome web browser and a multiple realms capable TURN server, quickly demonstrating feasibility.