This document updates RFC6265 by defining a SameSite attribute which allows servers to assert that a cookie ought not to be sent along with cross-site requests. This assertion allows user agents to mitigate the risk of cross-origin information leakage, and provides some protection against cross-site request forgery attacks.
Discussion of this draft takes place on the HTTP working group mailing list (firstname.lastname@example.org), which is archived at https://lists.w3.org/Archives/Public/ietf-http-wg/.
Working Group information can be found at http://httpwg.github.io/; source code and issues list for this draft can be found at https://github.com/httpwg/http-extensions/labels/cookie-same-site.
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 April 4, 2018.
Copyright © 2017 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.
Section 8.2 of [RFC6265] eloquently notes that cookies may be employed as a form of ambient authority, attached by default to requests the user agent sends on a user’s behalf. Even when an attacker doesn’t know the contents of a user’s cookies, she can still execute commands on the user’s behalf (and with the user’s authority) by asking the user agent to send HTTP requests to unwary servers. These malicious requests will include any of the user’s previously-set cookies, and therefore can be difficult to distinguish from benign requests on the user’s behalf.
Here, we update [RFC6265] with a simple mitigation strategy that allows servers to declare certain cookies as “same-site”, meaning they should not be attached to “cross-site” requests (as defined in section 2.1 of this specification).
Note that the mechanism outlined here is backwards compatible with the existing cookie syntax. Servers may serve these cookies to all user agents; those that do not support the SameSite attribute will simply store a cookie which is attached to all relevant requests, just as they do today.
Same-site cookies are intended to provide a solid layer of defense-in-depth against attacks which require embedding an authenticated request into an attacker-controlled context:
- Timing attacks which yield cross-origin information leakage (such as those detailed in [pixel-perfect]) can be substantially mitigated by setting the SameSite attribute on authentication cookies. The attacker will only be able to embed unauthenticated resources, as embedding mechanisms such as <iframe> will yield cross-site requests.
- Cross-site script inclusion (XSSI) attacks are likewise mitigated by setting the SameSite attribute on authentication cookies. The attacker will not be able to include authenticated resources via <script> or <link>, as these embedding mechanisms will likewise yield cross-site requests.
- Cross-site request forgery (CSRF) attacks which rely on top-level navigation (HTML <form> POSTs, for instance) can also be mitigated by treating these navigational requests as “cross-site”.
- Same-site cookies have some marginal value for policy or regulatory purposes, as cookies which are not delivered with cross-site requests cannot be directly used for tracking purposes. It may be valuable for an origin to assert that its cookies should not be sent along with cross-site requests in order to limit its exposure to non-technical risk.
Same-site cookies are set via the SameSite attribute in the Set-Cookie header field. That is, given a server’s response to a user agent which contains the following header field:
Set-Cookie: SID=31d4d96e407aad42; SameSite=Strict
Subsequent requests from that user agent can be expected to contain the following header field if and only if both the requested resource and the resource in the top-level browsing context match the cookie.
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 [RFC2119].
This specification uses the Augmented Backus-Naur Form (ABNF) notation of [RFC5234].
Two sequences of octets are said to case-insensitively match each other if and only if they are equivalent under the i;ascii-casemap collation defined in [RFC4790].
The terms “active document”, “ancestor browsing context”, “browsing context”, “dedicated worker”, “Document”, “WorkerGlobalScope”, “sandboxed origin browsing context flag”, “parent browsing context”, “shared worker”, “the worker’s Documents”, “nested browsing context”, and “top-level browsing context” are defined in [HTML].
“Service Workers” are defined in the Service Workers specification [SERVICE-WORKERS].
The term “origin”, the mechanism of deriving an origin from a URI, and the “the same” matching algorithm for origins are defined in [RFC6454].
“Safe” HTTP methods include GET, HEAD, OPTIONS, and TRACE, as defined in Section 4.2.1 of [RFC7231].
The term “public suffix” is defined in a note in Section 5.3 of [RFC6265] as “a domain that is controlled by a public registry”, and are also know as “effective top-level domains” (eTLDs). For example, example.com’s public suffix is com. User agents SHOULD use an up-to-date public suffix list, such as the one maintained by Mozilla at [PSL].
An origin’s “registered domain” is the origin’s host’s public suffix plus the label to its left. That is, for https://www.example.com, the public suffix is com, and the registered domain is example.com. This concept is defined more rigorously in [PSL], and is also know as “effective top-level domain plus one” (eTLD+1).
The term “request”, as well as a request’s “client”, “current url”, “method”, and “target browsing context”, are defined in [FETCH].
A request is “same-site” if its target’s URI’s origin’s registered domain is an exact match for the request’s client’s “site for cookies” or if the request has no client, and “cross-site” otherwise. To be more precise, for a given request (“request”), the following algorithm returns same-site or cross-site:
- If request’s client is null, return same-site.
- Let site be request’s client’s “site for cookies” (as defined in the following sections).
- Let target be the registered domain of request’s current url.
- If site is an exact match for target, return same-site.
- Return cross-site.
The URI displayed in a user agent’s address bar is the only security context directly exposed to users, and therefore the only signal users can reasonably rely upon to determine whether or not they trust a particular website. The registered domain of that URI’s origin represents the context in which a user most likely believes themselves to be interacting. We’ll label this domain the “top-level site”.
For a document displayed in a top-level browsing context, we can stop here: the document’s “site for cookies” is the top-level site.
For documents which are displayed in nested browsing contexts, we need to audit the origins of each of a document’s ancestor browsing contexts’ active documents in order to account for the “multiple-nested scenarios” described in Section 4 of [RFC7034]. These document’s “site for cookies” is the top-level site if and only if the document and each of its ancestor documents’ origins have the same registered domain as the top-level site. Otherwise its “site for cookies” is the empty string.
Given a Document (document), the following algorithm returns its “site for cookies” (either a registered domain, or the empty string):
- Let top-document be the active document in document’s browsing context’s top-level browsing context.
- Let top-origin be the origin of top-document’s URI if top-document’s sandboxed origin browsing context flag is set, and top-document’s origin otherwise.
- Let documents be a list containing document and each of document’s ancestor browsing contexts’ active documents.
- For each item in documents:
- Let origin be the origin of item’s URI if item’s sandboxed origin browsing context flag is set, and item’s origin otherwise.
- If origin’s host’s registered domain is not an exact match for top-origin’s host’s registered domain, return the empty string.
- Return top-site.
Worker-driven requests aren’t as clear-cut as document-driven requests, as there isn’t a clear link between a top-level browsing context and a worker. This is especially true for Service Workers [SERVICE-WORKERS], which may execute code in the background, without any document visible at all.
Note: The descriptions below assume that workers must be same-origin with the documents that instantiate them. If this invariant changes, we’ll need to take the worker’s script’s URI into account when determining their status.
Service Workers are more complicated, as they act as a completely separate execution context with only tangential relationship to the Document which registered them.
Requests which simply pass through a service worker will be handled as described above: the request’s client will be the Document or Worker which initiated the request, and its “site for cookies” will be those defined in Section 2.1.1 and Section 126.96.36.199
Requests which are initiated by the Service Worker itself (via a direct call to fetch(), for instance), on the other hand, will have a client which is a ServiceWorkerGlobalScope. Its “site for cookies” will be the registered domain of the Service Worker’s URI.
Given a ServiceWorkerGlobalScope (worker), the following algorithm returns its “site for cookies” (either a registered domain, or the empty string):
- Return worker’s origin’s host’s registered domain.
This section describes extensions to [RFC6265] necessary to implement the server-side requirements of the SameSite attribute.
Add SameSite to the list of accepted attributes in the Set-Cookie header field’s value by replacing the cookie-av token definition in Section 4.1.1 of [RFC6265] with the following ABNF grammar:
cookie-av = expires-av / max-age-av / domain-av / path-av / secure-av / httponly-av / samesite-av / extension-av samesite-av = "SameSite=" samesite-value samesite-value = "Strict" / "Lax"
The “SameSite” attribute limits the scope of the cookie such that it will only be attached to requests if those requests are same-site, as defined by the algorithm in Section 2.1. For example, requests for https://example.com/sekrit-image will attach same-site cookies if and only if initiated from a context whose “site for cookies” is “example.com”.
If the “SameSite” attribute’s value is “Strict”, the cookie will only be sent along with “same-site” requests. If the value is “Lax”, the cookie will be sent with same-site requests, and with “cross-site” top-level navigations, as described in Section 4.1.1. If the “SameSite” attribute’s value is neither of these, the cookie will be ignored.
The changes to the Cookie header field suggested in Section 4.3 provide additional detail.
This section describes extensions to [RFC6265] necessary in order to implement the client-side requirements of the SameSite attribute.
The following attribute definition should be considered part of the the Set-Cookie algorithm as described in Section 5.2 of [RFC6265]:
If the attribute-name case-insensitively matches the string “SameSite”, the user agent MUST process the cookie-av as follows:
- If cookie-av’s attribute-value is not a case-insensitive match for “Strict” or “Lax”, ignore the cookie-av.
- Let enforcement be “Lax” if cookie-av’s attribute-value is a case-insensitive match for “Lax”, and “Strict” otherwise.
- Append an attribute to the cookie-attribute-list with an attribute-name of “SameSite” and an attribute-value of enforcement.
Same-site cookies in “Strict” enforcement mode will not be sent along with top-level navigations which are triggered from a cross-site document context. As discussed in Section 5.2, this might or might not be compatible with existing session management systems. In the interests of providing a drop-in mechanism that mitigates the risk of CSRF attacks, developers may set the SameSite attribute in a “Lax” enforcement mode that carves out an exception which sends same-site cookies along with cross-site requests if and only if they are top-level navigations which use a “safe” (in the [RFC7231] sense) HTTP method.
Lax enforcement provides reasonable defense in depth against CSRF attacks that rely on unsafe HTTP methods (like POST), but does not offer a robust defense against CSRF as a general category of attack:
- Attackers can still pop up new windows or trigger top-level navigations in order to create a “same-site” request (as described in section 2.1), which is only a speedbump along the road to exploitation.
- Features like <link rel='prerender'> [prerendering] can be exploited to create “same-site” requests without the risk of user detection.
When possible, developers should use a session management mechanism such as that described in Section 5.2 to mitigate the risk of CSRF more completely.
Note: There’s got to be a better way to specify this. Until I figure out what that is, monkey-patching!
Alter Section 5.3 of [RFC6265] as follows:
- Add samesite-flag to the list of each cookie’s fields defined in the first paragraph. Note: this field’s value is one of “None”, “Strict”, or “Lax”.
- Before step 11 of the current algorithm, add the following:
- If the cookie-attribute-list contains an attribute with an attribute-name of “SameSite”, set the cookie’s samesite-flag to attribute-value (“Strict” or “Lax”). Otherwise, set the cookie’s samesite-flag to “None”.
- If the cookie’s samesite-flag is not “None”, and the request which generated the cookie’s client’s “site for cookies” is not an exact match for request-uri’s host’s registered domain, then abort these steps and ignore the newly created cookie entirely.
Same-site cookies in and of themselves don’t do anything to address the general privacy concerns outlined in Section 7.1 of [RFC6265]. The SameSite attribute is set by the server, and serves to mitigate the risk of certain kinds of attacks that the server is worried about. The user is not involved in this decision. Moreover, a number of side-channels exist which could allow a server to link distinct requests even in the absence of cookies. Connection and/or socket pooling, Token Binding, and Channel ID all offer explicit methods of identification that servers could take advantage of.
As outlined in [RFC7258], pervasive monitoring is an attack. Cookies play a large part in enabling such monitoring, as they are responsible for maintaining state in HTTP connections. We considered restricting same-site cookies to secure contexts [secure-contexts] as a mitigation but decided against doing so, as same-site cookies should result in a strict reduction in the number of cookies floating around in cross-site contexts. That is, even if http://not-example.com embeds a resource from http://example.com/, that resource will not be “same-site”, and http://example.com’s cookies simply cannot be used to correlate user behavior across distinct origins.
7.1. Normative References
- van Kesteren, A., “Fetch”, n.d., <https://fetch.spec.whatwg.org/>.
- Hickson, I., Pieters, S., van Kesteren, A., Jägenstedt, P., and D. Denicola, “HTML”, n.d., <https://html.spec.whatwg.org/>.
- “Public Suffix List”, n.d., <https://publicsuffix.org/list/>.
- Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels”, BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, <https://www.rfc-editor.org/info/rfc2119>.
- Newman, C., Duerst, M., and A. Gulbrandsen, “Internet Application Protocol Collation Registry”, RFC 4790, DOI 10.17487/RFC4790, March 2007, <https://www.rfc-editor.org/info/rfc4790>.
- Crocker, D., Ed. and P. Overell, “Augmented BNF for Syntax Specifications: ABNF”, STD 68, RFC 5234, DOI 10.17487/RFC5234, January 2008, <https://www.rfc-editor.org/info/rfc5234>.
- Barth, A., “HTTP State Management Mechanism”, RFC 6265, DOI 10.17487/RFC6265, April 2011, <https://www.rfc-editor.org/info/rfc6265>.
- Barth, A., “The Web Origin Concept”, RFC 6454, DOI 10.17487/RFC6454, December 2011, <https://www.rfc-editor.org/info/rfc6454>.
- Fielding, R., Ed. and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Semantics and Content”, RFC 7231, DOI 10.17487/RFC7231, June 2014, <https://www.rfc-editor.org/info/rfc7231>.
- Farrell, S. and H. Tschofenig, “Pervasive Monitoring Is an Attack”, BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 2014, <https://www.rfc-editor.org/info/rfc7258>.
- Russell, A., Song, J., and J. Archibald, “Service Workers”, n.d., <http://www.w3.org/TR/service-workers/>.
7.2. Informative References
- Ross, D. and T. Gondrom, “HTTP Header Field X-Frame-Options”, RFC 7034, DOI 10.17487/RFC7034, October 2013, <https://www.rfc-editor.org/info/rfc7034>.
- Chen, E., Bau, J., Reis, C., Barth, A., and C. Jackson, “App Isolation - Get the Security of Multiple Browsers with Just One”, 2011, <http://www.collinjackson.com/research/papers/appisolation.pdf>.
- Stone, P., “Pixel Perfect Timing Attacks with HTML5”, n.d., <http://www.contextis.com/documents/2/Browser_Timing_Attacks.pdf>.
- Bentzel, C., “Chrome Prerendering”, n.d., <https://www.chromium.org/developers/design-documents/prerender>.
- Goodwin, M. and J. Walker, “SameDomain Cookie Flag”, 2011, <http://people.mozilla.org/~mgoodwin/SameDomain/samedomain-latest.txt>.
- West, M., “Secure Contexts”, n.d., <https://w3c.github.io/webappsec-secure-contexts/>.
The same-site cookie concept documented here is indebited to Mark Goodwin’s and Joe Walker’s [samedomain-cookies]. Michal Zalewski, Artur Janc, Ryan Sleevi, Adam Barth, and Jeff Hodges provided particularly valuable feedback on this document.