Google, Inc

mkwst@google.com https://mikewest.org/

Mozilla

mgoodwin@mozilla.com https://www.computerist.org/

Applications and Real-Time HTTP Cookie 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 (ietf-http-wg@w3.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.

Section 8.2 of eloquently notes that cookies are 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. Here, we update 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). 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.

These 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 ) 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:

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 . This specification uses the Augmented Backus-Naur Form (ABNF) notation of . 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 . The terms “active document”, “ancestor browsing context”, “browsing context”, “document”, “WorkerGlobalScope”, “sandboxed origin browsing context flag”, “parent browsing context”, “the worker’s Documents”, “nested browsing context”, and “top-level browsing context” are defined in . “Service Workers” are defined in the Service Workers specification . The term “origin”, the mechanism of deriving an origin from a URI, and the “the same” matching algorithm for origins are defined in . “Safe” HTTP methods include GET, HEAD, OPTIONS, and TRACE, as defined in Section 4.2.1 of . The term “public suffix” is defined in a note in Section 5.3 of as “a domain that is controlled by a public registry”. 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 . An origin’s “registrable domain” is the origin’s host’s public suffix plus the label to its left. That is, https://www.example.com’s registrable domain is example.com. This concept is defined more rigorously in . The term “request”, as well as a request’s “client”, “current url”, “method”, and “target browsing context”, are defined in .

A request is “same-site” if its target’s URI’s origin’s registrable domain is an exact match for the request’s initiator’s “site for cookies”, 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 registrable 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 registrable 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 . 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 registrable 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 registrable 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 registrable domain is not an exact match for top-origin’s host’s registrable 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 , 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.

Dedicated workers are simple, as each dedicated worker is bound to one and only one document. Requests generated from a dedicated worker (via importScripts, XMLHttpRequest, fetch(), etc) define their “site for cookies” as that document’s “site for cookies”. Shared workers may be bound to multiple documents at once. As it is quite possible for those documents to have distinct “site for cookie” values, the worker’s “site for cookies” will be the empty string in cases where the values diverge, and the shared value in cases where the values agree. Given a WorkerGlobalScope (worker), the following algorithm returns its “site for cookies” (either a registrable domain, or the empty string): Let site be worker’s origin’s host’s registrable domain. For each document in worker’s Documents: Let document-site be document’s “site for cookies” (as defined in ). If document-site is not an exact match for site, return the empty string. Return site.

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 and 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 registrable domain of the Service Worker’s URI. Given a ServiceWorkerGlobalScope (worker), the following algorithm returns its “site for cookies” (either a registrable domain, or the empty string): Return worker’s origin’s host’s registrable domain.

This section describes extensions to 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 with the following ABNF grammar:

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 . 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”, or if the value is invalid, 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 . The changes to the Cookie header field suggested in provide additional detail.

This section describes extensions to 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 : 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.

By default, same-site cookies will not be sent along with top-level navigations. As discussed in , 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 sense) HTTP method. Lax enforcement provides reasonable defense in depth against CSRF attacks that rely on unsafe HTTP methods (like POST), but do 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'> 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 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 as follows: Add samesite-flag to the list of fields stored for each cookie. 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 registrable domain, then abort these steps and ignore the newly created cookie entirely.

Note: There’s got to be a better way to specify this. Until I figure out what that is, monkey-patching! Alter Section 5.4 of as follows: Add the following requirement to the list in step 1: If the cookie’s samesite-flag is not “None”, and the HTTP request is cross-site (as defined in then exclude the cookie unless all of the following statements hold: samesite-flag is “Lax” The HTTP request’s method is “safe”. The HTTP request’s target browsing context is a top-level browsing context. Note that the modifications suggested here concern themselves only with the “site for cookies” of the request’s client, and the registrable domain of the resource being requested. The cookie’s domain, path, and secure attributes do not come into play for these comparisons.

“SameSite” cookies offer a robust defense against CSRF attack when deployed in strict mode, and when supported by the client. It is, however, prudent to ensure that this designation is not the extent of a site’s defense against CSRF, as same-site navigations and submissions can certainly be executed in conjunction with other attack vectors such as cross-site scripting. Developers are strongly encouraged to deploy the usual server-side defenses (CSRF tokens, ensuring that “safe” HTTP methods are idempotent, etc) to mitigate the risk more fully. Additionally, client-side techniques such as those described in may also prove effective against CSRF, and are certainly worth exploring in combination with “SameSite” cookies.

Setting the SameSite attribute in “strict” mode provides robust defense in depth against CSRF attacks, but has the potential to confuse users unless sites’ developers carefully ensure that their session management systems deal reasonably well with top-level navigations. Consider the scenario in which a user reads their email at MegaCorp Inc’s webmail provider https://example.com/. They might expect that clicking on an emailed link to https://projects.com/secret/project would show them the secret project that they’re authorized to see, but if projects.com has marked their session cookies as SameSite, then this cross-site navigation won’t send them along with the request. projects.com will render a 404 error to avoid leaking secret information, and the user will be quite confused. Developers can avoid this confusion by adopting a session management system that relies on not one, but two cookies: one conceptualy granting “read” access, another granting “write” access. The latter could be marked as SameSite, and its absence would provide a reauthentication step before executing any non-idempotent action. The former could drop the SameSite attribute entirely, or choose the “Lax” version of enforcement, in order to allow users access to data via top-level navigation.

The SameSite attribute is inappropriate for some important use-cases. In particular, note that content intended for embedding in a cross-site contexts (social networking widgets or commenting services, for instance) will not have access to such cookies. Cross-site cookies may be required in order to provide seamless functionality that relies on a user’s state. Likewise, some forms of Single-Sign-On might require authentication in a cross-site context; these mechanisms will not function as intended with same-site cookies.

Same-site cookies in and of themselves don’t do anything to address the general privacy concerns outlined in Section 7.1 of . The 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 , 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 as a mitigation but decided against doing so, as this feature 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.

Mozilla Google, Inc. Opera Mozilla Opera Google, Inc. In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. Many Internet application protocols include string-based lookup, searching, or sorting operations. However, the problem space for searching and sorting international strings is large, not fully explored, and is outside the area of expertise for the Internet Engineering Task Force (IETF). Rather than attempt to solve such a large problem, this specification creates an abstraction framework so that application protocols can precisely identify a comparison function, and the repertoire of comparison functions can be extended in the future. [STANDARDS-TRACK] Internet technical specifications often need to define a formal syntax. Over the years, a modified version of Backus-Naur Form (BNF), called Augmented BNF (ABNF), has been popular among many Internet specifications. The current specification documents ABNF. It balances compactness and simplicity with reasonable representational power. The differences between standard BNF and ABNF involve naming rules, repetition, alternatives, order-independence, and value ranges. This specification also supplies additional rule definitions and encoding for a core lexical analyzer of the type common to several Internet specifications. [STANDARDS-TRACK] This document defines the HTTP Cookie and Set-Cookie header fields. These header fields can be used by HTTP servers to store state (called cookies) at HTTP user agents, letting the servers maintain a stateful session over the mostly stateless HTTP protocol. Although cookies have many historical infelicities that degrade their security and privacy, the Cookie and Set-Cookie header fields are widely used on the Internet. This document obsoletes RFC 2965. [STANDARDS-TRACK] This document defines the concept of an "origin", which is often used as the scope of authority or privilege by user agents. Typically, user agents isolate content retrieved from different origins to prevent malicious web site operators from interfering with the operation of benign web sites. In addition to outlining the principles that underlie the concept of origin, this document details how to determine the origin of a URI and how to serialize an origin into a string. It also defines an HTTP header field, named "Origin", that indicates which origins are associated with an HTTP request. [STANDARDS-TRACK] The Hypertext Transfer Protocol (HTTP) is a stateless \%application- level protocol for distributed, collaborative, hypertext information systems. This document defines the semantics of HTTP/1.1 messages, as expressed by request methods, request header fields, response status codes, and response header fields, along with the payload of messages (metadata and body content) and mechanisms for content negotiation. Pervasive monitoring is a technical attack that should be mitigated in the design of IETF protocols, where possible. To improve the protection of web applications against clickjacking, this document describes the X-Frame-Options HTTP header field, which declares a policy, communicated from the server to the client browser, regarding whether the browser may display the transmitted content in frames that are part of other web pages.

The same-site cookie concept documented here is indebited to Mark Goodwin’s and Joe Walker’s . Michal Zalewski, Artur Janc, Ryan Sleevi, and Adam Barth provided particularly valuable feedback on this document.