| OAuth | W. Denniss |
| Internet-Draft | |
| Intended status: Standards Track | J. Bradley |
| Expires: September 12, 2019 | Ping Identity |
| M. Jones | |
| Microsoft | |
| H. Tschofenig | |
| ARM Limited | |
| March 11, 2019 |
OAuth 2.0 Device Authorization Grant
draft-ietf-oauth-device-flow-15
The OAuth 2.0 Device Authorization Grant is designed for internet-connected devices that either lack a browser to perform a user-agent based authorization, or are input-constrained to the extent that requiring the user to input text in order to authenticate during the authorization flow is impractical. It enables OAuth clients on such devices (like smart TVs, media consoles, digital picture frames, and printers) to obtain user authorization to access protected resources without using an on-device user-agent.
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 https://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 September 12, 2019.
Copyright (c) 2019 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 (https://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.
This OAuth 2.0 protocol extension, sometimes referred to as "device flow", enables OAuth clients to request user authorization from applications on devices that have limited input capabilities or lack a suitable browser. Such devices include those smart TVs, media console, picture frames and printers which lack an easy input method or suitable browser required for traditional OAuth interactions. The authorization flow defined by this specification instructs the user to review the authorization request on a secondary device, such as a smartphone which does have the requisite input and browser capabilities to complete the user interaction.
The Device Authorization Grant is not intended to replace browser-based OAuth in native apps on capable devices like smartphones. Those apps should follow the practices specified in OAuth 2.0 for Native Apps.
The operating requirements to be able to use this authorization grant type are:
As the device authorization grant does not require two-way communication between the OAuth client and the user-agent (unlike other OAuth 2 grant types such as the Authorization Code and Implicit grant types), it supports several use cases that cannot be served by those other approaches.
Instead of interacting with the end user's user agent, the client instructs the end user to use another computer or device and connect to the authorization server to approve the access request. Since the protocol supports clients that can't receive incoming requests, clients poll the authorization server repeatedly until the end user completes the approval process.
The device typically chooses the set of authorization servers to support (i.e., its own authorization server, or those by providers it has relationships with). It is not uncommon for the device application to support only a single authorization server, such as with a TV application for a specific media provider that supports only that media provider's authorization server. The user may not have an established relationship yet with that authorization provider, though one can potentially be set up during the authorization flow.
+----------+ +----------------+
| |>---(A)-- Client Identifier --->| |
| | | |
| |<---(B)-- Device Code, ---<| |
| | User Code, | |
| Device | & Verification URI | |
| Client | | |
| | [polling] | |
| |>---(E)-- Device Code, --->| |
| | & Client Identifier | |
| | | Authorization |
| |<---(F)-- Access Token ---<| Server |
+----------+ (& Optional Refresh Token) | |
v | |
: | |
(C) User Code & Verification URI | |
: | |
v | |
+----------+ | |
| End user | | |
| at |<---(D)-- End user reviews --->| |
| Browser | authorization request | |
+----------+ +----------------+
Figure 1: Device Authorization Flow
The device authorization flow illustrated in Figure 1 includes the following steps:
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.
This specification defines a new OAuth endpoint, the device authorization endpoint. This is separate from the OAuth authorization endpoint defined in [RFC6749] with which the user interacts with via a user-agent (i.e., a browser). By comparison, when using the device authorization endpoint, the OAuth client on the device interacts with the authorization server directly without presenting the request in a user-agent, and the end user authorizes the request on a separate device. This interaction is defined as follows.
The client initiates the authorization flow by requesting a set of verification codes from the authorization server by making an HTTP "POST" request to the device authorization endpoint.
The client constructs the request with the following parameters, sent as the body of the request, encoded with the application/x-www-form-urlencoded encoding algorithm defined by Section 4.10.22.6 of [HTML5]:
For example, the client makes the following HTTPS request:
POST /device_authorization HTTP/1.1 Host: server.example.com Content-Type: application/x-www-form-urlencoded client_id=459691054427
All requests from the device MUST use the Transport Layer Security (TLS) [RFC8446] protocol and implement the best practices of BCP 195 [RFC7525].
Parameters sent without a value MUST be treated as if they were omitted from the request. The authorization server MUST ignore unrecognized request parameters. Request and response parameters MUST NOT be included more than once.
The client authentication requirements of Section 3.2.1 of [RFC6749] apply to requests on this endpoint, which means that confidential clients (those that have established client credentials) authenticate in the same manner as when making requests to the token endpoint, and public clients provide the "client_id" parameter to identify themselves.
Due to the polling nature of this protocol (as specified in Section 3.4), care is needed to avoid overloading the capacity of the token endpoint. To avoid unneeded requests on the token endpoint, the client SHOULD only commence a device authorization request when prompted by the user, and not automatically, such as when the app starts or when the previous authorization session expires or fails.
In response, the authorization server generates a unique device verification code and an end-user code that are valid for a limited time and includes them in the HTTP response body using the "application/json" format [RFC8259] with a 200 (OK) status code. The response contains the following parameters:
For example:
HTTP/1.1 200 OK
Content-Type: application/json
Cache-Control: no-store
{
"device_code": "GmRhmhcxhwAzkoEqiMEg_DnyEysNkuNhszIySk9eS",
"user_code": "WDJB-MJHT",
"verification_uri": "https://example.com/device",
"verification_uri_complete":
"https://example.com/device?user_code=WDJB-MJHT",
"expires_in": 1800,
"interval": 5
}
In the event of an error (such as an invalidly configured client), the authorization server responds in the same way as the token endpoint specified in Section 5.2 of [RFC6749].
After receiving a successful Authorization Response, the client displays or otherwise communicates the user_code and the verification_uri to the end user and instructs them to visit the URI in a user agent on a secondary device (for example, in a browser on their mobile phone), and enter the user code.
+-----------------------------------------------+
| |
| Using a browser on another device, visit: |
| https://example.com/device |
| |
| And enter the code: |
| WDJB-MJHT |
| |
+-----------------------------------------------+
Figure 2: Example User Instruction
The authorizing user navigates to the verification_uri and authenticates with the authorization server in a secure TLS-protected ([RFC8446]) session. The authorization server prompts the end user to identify the device authorization session by entering the user_code provided by the client. The authorization server should then inform the user about the action they are undertaking and ask them to approve or deny the request. Once the user interaction is complete, the server MAY inform the user to return to their device.
During the user interaction, the device continuously polls the token endpoint with the device_code, as detailed in Section 3.4, until the user completes the interaction, the code expires, or another error occurs. The device_code is not intended for the end user directly, and thus should not be displayed during the interaction to avoid confusing the end user.
Authorization servers supporting this specification MUST implement a user interaction sequence that starts with the user navigating to verification_uri and continues with them supplying the user_code at some stage during the interaction. Other than that, the exact sequence and implementation of the user interaction is up to the authorization server, for example, the authorization server may enable new users to sign up for an account during the authorization flow, or add additional security verification steps.
It is NOT RECOMMENDED for authorization servers to include the user code in the verification URI (verification_uri), as this increases the length and complexity of the URI that the user must type. While the user must still type the same number of characters with the user_code separated, once they successfully navigate to the verification_uri, any errors in entering the code can be highlighted by the authorization server to improve the user experience. The next section documents user interaction with verification_uri_complete, which is designed to carry both pieces of information.
When verification_uri_complete is included in the Authorization Response (Section 3.2), clients MAY present this URI in a non-textual manner using any method that results in the browser being opened with the URI, such as with QR (Quick Response) codes or NFC (Near Field Communication), to save the user typing the URI.
For usability reasons, it is RECOMMENDED for clients to still display the textual verification URI (verification_uri) for users not able to use such a shortcut. Clients MUST still display the user_code, as the authorization server will require the user to confirm it to disambiguate devices, or as a remote phishing mitigation (See Section 5.4).
If the user starts the user interaction by browsing to verification_uri_complete, then the user interaction described in Section 3.3 is still followed, but with the optimization that the user does not need to type the user_code. The server SHOULD display the user_code to the user and ask them to verify that it matches the user_code being displayed on the device, to confirm they are authorizing the correct device. As before, in addition to taking steps to confirm the identity of the device, the user should also be afforded the choice to approve or deny the authorization request.
+-------------------------------------------------+
| |
| Scan the QR code, or using +------------+ |
| a browser on another device, |[_].. . [_]| |
| visit: | . .. . .| |
| https://example.com/device | . . . ....| |
| |. . . . | |
| And enter the code: |[_]. ... . | |
| WDJB-MJHT +------------+ |
| |
+-------------------------------------------------+
Figure 3: Example User Instruction with QR Code Representation of the Complete Verification URI
After displaying instructions to the user, the client makes an Access Token Request to the token endpoint (as defined by Section 3.2 of [RFC6749]) with a grant_type of
urn:ietf:params:oauth:grant-type:device_code. This is an extension grant type (as defined by Section 4.5 of [RFC6749]) created by this specification, with the following parameters:
For example, the client makes the following HTTPS request (line breaks are for display purposes only):
POST /token HTTP/1.1 Host: server.example.com Content-Type: application/x-www-form-urlencoded grant_type=urn%3Aietf%3Aparams%3Aoauth%3Agrant-type%3Adevice_code &device_code=GmRhmhcxhwAzkoEqiMEg_DnyEysNkuNhszIySk9eS &client_id=459691054427
If the client was issued client credentials (or assigned other authentication requirements), the client MUST authenticate with the authorization server as described in Section 3.2.1 of [RFC6749]. Note that there are security implications of statically distributed client credentials, see Section 5.6.
The response to this request is defined in Section 3.5. Unlike other OAuth grant types, it is expected for the client to try the Access Token Request repeatedly in a polling fashion, based on the error code in the response.
If the user has approved the grant, the token endpoint responds with a success response defined in Section 5.1 of [RFC6749]; otherwise it responds with an error, as defined in Section 5.2 of [RFC6749].
In addition to the error codes defined in Section 5.2 of [RFC6749], the following error codes are specified for use with the device authorization grant in token endpoint responses:
The authorization_pending and slow_down error codes define particularly unique behavior, as they indicate that the OAuth client should continue to poll the token endpoint by repeating the token request (implementing the precise behavior defined above). If the client receives an error response with any other error code, it MUST stop polling and SHOULD react accordingly, for example, by displaying an error to the user.
On encountering a connection timeout, clients MUST unilaterally reduce their polling frequency before retrying. The use of an exponential backoff algorithm to achieve this, such as by doubling the polling interval on each such connection timeout, is RECOMMENDED.
The assumption of this specification is that the separate device the user is authorizing the request on does not have a way to communicate back to device with the OAuth client. This protocol only requires a one-way channel in order to maximise the viability of the protocol in restricted environments, like an application running on a TV that is only capable of outbound requests. If a return channel were to exist for the chosen user interaction interface, then the device MAY wait until notified on that channel that the user has completed the action before initiating the token request (as an alternative to polling). Such behavior is, however, outside the scope of this specification.
Support for this specification MAY be declared in the OAuth 2.0 Authorization Server Metadata [RFC8414] by including the value urn:ietf:params:oauth:grant-type:device_code in the grant_types_supported parameter, and by adding the following new parameter:
Since the user code is typed by the user, shorter codes are more desirable for usability reasons. This means the entropy is typically less than would be used for the device code or other OAuth bearer token types where the code length does not impact usability. It is therefore recommended that the server rate-limit user code attempts.
The user code SHOULD have enough entropy that when combined with rate limiting and other mitigations makes a brute-force attack infeasible. For example, it's generally held that 128-bit symmetric keys for encryption are seen as good enough today because an attacker has to put in 2^96 work to have a 2^-32 chance of guessing correctly via brute force. The rate limiting and finite lifetime on the user code places an artificial limit on the amount of work an attacker can "do", so if, for instance, one uses a 8-character base-20 user code (with roughly 34.5 bits of entropy), the rate-limiting interval and validity period would need to only allow 5 attempts in order to get the same 2^-32 probability of success by random guessing.
A successful brute forcing of the user code would enable the attacker to authenticate with their own credentials and make an authorization grant to the device. This is the opposite scenario to an OAuth bearer token being brute forced, whereby the attacker gains control of the victim's authorization grant. Such attacks may not always make economic sense, for example for a video app the device owner may then be able to purchase movies using the attacker's account, though a privacy risk would still remain and thus is important to protect against. Furthermore, some uses of the device flow give the granting account the ability to perform actions such as controlling the device, which needs to be protected.
The precise length of the user code and the entropy contained within is at the discretion of the authorization server, which needs to consider the sensitivity of their specific protected resources, the practicality of the code length from a usability standpoint, and any mitigations that are in place such as rate-limiting, when determining the user code format.
An attacker who guesses the device code would be able to potentially obtain the authorization code once the user completes the flow. As the device code is not displayed to the user and thus there are no usability considerations on the length, a very high entropy code SHOULD be used.
Unlike other native application OAuth 2.0 flows, the device requesting the authorization is not the same as the device that the user grants access from. Thus, signals from the approving user's session and device are not relevant to the trustworthiness of the client device.
Note that if an authorization server used with this flow is malicious, then it could man-in-the-middle the backchannel flow to another authorization server. In this scenario, the man-in-the-middle is not completely hidden from sight, as the end user would end up on the authorization page of the wrong service, giving them an opportunity to notice that the URL in the browser's address bar is wrong. For this to be possible, the device manufacturer must either directly be the attacker, shipping a device intended to perform the man-in-the-middle attack, or be using an authorization server that is controlled by an attacker, possibly because the attacker compromised the authorization server used by the device. In part, the person purchasing the device is counting on it and its business partners to be trustworthy.
It is possible for the device flow to be initiated on a device in an attacker's possession. For example, an attacker might send an email instructing the target user to visit the verification URL and enter the user code. To mitigate such an attack, it is RECOMMENDED to inform the user that they are authorizing a device during the user interaction step (see Section 3.3), and to confirm that the device is in their possession. The authorization server SHOULD display information about the device so that the person can notice if a software client was attempting to impersonating a hardware device.
For authorization servers that support the option specified in Section 3.3.1 for the client to append the user code to the authorization URI, it is particularly important to confirm that the device is in the user's possession, as the user no longer has to type the code manually. One possibility is to display the code during the authorization flow and asking the user to verify that the same code is being displayed on the device they are setting up.
The user code needs to have a long enough lifetime to be useable (allowing the user to retrieve their secondary device, navigate to the verification URI, login, etc.), but should be sufficiently short to limit the usability of a code obtained for phishing. This doesn't prevent a phisher presenting a fresh token, particularly in the case they are interacting with the user in real time, but it does limit the viability of codes sent over email or SMS.
While the device is pending authorization, it may be possible for a malicious user to physically spy on the device user interface (by viewing the screen on which it's displayed, for example) and hijack the session by completing the authorization faster than the user that initiated it. Devices SHOULD take into account the operating environment when considering how to communicate the code to the user to reduce the chances it will be observed by a malicious user.
Device clients are generally incapable of maintaining the confidentiality of their credentials, as users in possession of the device can reverse engineer it and extract the credentials. Therefore, unless additional measures are taken, they should be treated as public clients (as defined by Section 2.1 of OAuth 2.0) susceptible to impersonation. The security considerations of Section 5.3.1 of [RFC6819] and Sections 8.5 and 8.6 of [RFC8252] apply to such clients.
The user may also be able to obtain the device_code and/or other OAuth bearer tokens issued to their client, which would allow them to use their own authorization grant directly by impersonating the client. Given that the user in possession of the client credentials can already impersonate the client and create a new authorization grant (with a new device_code), this doesn't represent a separate impersonation vector.
There is no requirement that the user code be displayed by the device visually. Other methods of one-way communication can potentially be used, such as text-to-speech audio, or Bluetooth Low Energy. To mitigate an attack in which a malicious user can bootstrap their credentials on a device not in their control, it is RECOMMENDED that any chosen communication channel only be accessible by people in close proximity. E.g., users who can see, or hear the device.
This section is a non-normative discussion of usability considerations.
For many users, their nearest Internet-connected device will be their mobile phone, and typically these devices offer input methods that are more time consuming than a computer keyboard to change the case or input numbers. To improve usability (improving entry speed, and reducing retries), these limitations should be taken into account when selecting the user-code character set.
One way to improve input speed is to restrict the character set to case-insensitive A-Z characters, with no digits. These characters can typically be entered on a mobile keyboard without using modifier keys. Further removing vowels to avoid randomly creating words results in the base-20 character set: BCDFGHJKLMNPQRSTVWXZ. Dashes or other punctuation may be included for readability.
An example user code following this guideline containing 8 significant characters and dashes added for end-user readability, with a resulting entropy of 20^8: WDJB-MJHT.
Pure numeric codes are also a good choice for usability, especially for clients targeting locales where A-Z character keyboards are not used, though their length needs to be longer to maintain a high entropy.
An example numeric user code containing 9 significant digits and dashes added for end-user readability, with an entropy of 10^9: 019-450-730.
When processing the inputted user code, the server should strip dashes and other punctuation it added for readability (making the inclusion of that punctuation by the user optional). For codes using only characters in the A-Z range as with the base-20 charset defined above, the user's input should be upper-cased before comparison to account for the fact that the user may input the equivalent lower-case characters. Further stripping of all characters outside the user_code charset is recommended to reduce instances where an errantly typed character (like a space character) invalidates otherwise valid input.
It is RECOMMENDED to avoid character sets that contain two or more characters that can easily be confused with each other like "0" and "O", or "1", "l" and "I". Furthermore, the extent practical, where a character set contains one character that may be confused with characters outside the character set the character outside the set MAY be substituted with the one in the character set that it is commonly confused with (for example, "O" for "0" when using a numerical 0-9 character set).
Devices and authorization servers MAY negotiate an alternative code transmission and user interaction method in addition to the one described in Section 3.3. Such an alternative user interaction flow could obviate the need for a browser and manual input of the code, for example, by using Bluetooth to transmit the code to the authorization server's companion app. Such interaction methods can utilize this protocol, as ultimately, the user just needs to identify the authorization session to the authorization server; however, user interaction other than via the verification URI is outside the scope of this specification.
This specification registers the following values in the IANA "OAuth Parameters" registry [IANA.OAuth.Parameters] established by [RFC6749].
This specification registers the following values in the IANA "OAuth URI" registry [IANA.OAuth.Parameters] established by [RFC6755].
This specification registers the following values in the IANA "OAuth Extensions Error Registry" registry [IANA.OAuth.Parameters] established by [RFC6749].
This specification registers the following values in the IANA "OAuth 2.0 Authorization Server Metadata" registry [IANA.OAuth.Parameters] established by [RFC8414].
The starting point for this document was the Internet-Draft draft-recordon-oauth-v2-device, authored by David Recordon and Brent Goldman, which itself was based on content in draft versions of the OAuth 2.0 protocol specification removed prior to publication due to a then lack of sufficient deployment expertise. Thank you to the OAuth working group members who contributed to those earlier drafts.
This document was produced in the OAuth working group under the chairpersonship of Rifaat Shekh-Yusef and Hannes Tschofenig with Benjamin Kaduk, Kathleen Moriarty, and Eric Rescorla serving as Security Area Directors.
The following individuals contributed ideas, feedback, and wording that shaped and formed the final specification:
Alissa Cooper, Ben Campbell, Brian Campbell, Roshni Chandrashekhar, Eric Fazendin, Benjamin Kaduk, Jamshid Khosravian, Torsten Lodderstedt, James Manger, Dan McNulty, Breno de Medeiros, Simon Moffatt, Stein Myrseth, Emond Papegaaij, Justin Richer, Adam Roach, Nat Sakimura, Andrew Sciberras, Marius Scurtescu, Filip Skokan, Ken Wang, and Steven E. Wright.
[[ to be removed by the RFC Editor before publication as an RFC ]]
-15
-14
-13
-12
-11
-10
-09
-08
-07
-06
-05
-04
-03
-02
-01
-00