HTTP Working Group | M. Nottingham |
Internet-Draft | M. Thomson |
Intended status: Experimental | Mozilla |
Expires: September 18, 2017 | March 17, 2017 |
This document describes how http URIs can be accessed using Transport Layer Security (TLS) and HTTP/2 to mitigate pervasive monitoring attacks. This mechanism not a replacement for https URIs; it is vulnerable to active attacks.¶
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 September 18, 2017.¶
Copyright (c) 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.¶
This document describes a use of HTTP Alternative Services [RFC7838] to decouple the URI scheme from the use and configuration of underlying encryption. It allows an http URI to be accessed using HTTP/2 [RFC7230] and Transport Layer Security (TLS) [RFC5246] with Opportunistic Security [RFC7435].¶
This document describes a usage model whereby sites can serve http URIs over TLS, thereby avoiding the problem of serving Mixed Content (described in [W3C.CR-mixed-content-20160802]) while still providing protection against passive attacks.¶
Opportunistic Security does not provide the same guarantees as using TLS with https URIs, because it is vulnerable to active attacks, and does not change the security context of the connection. Normally, users will not be able to tell that it is in use (i.e., there will be no “lock icon”).¶
The immediate goal is to make the use of HTTP more robust in the face of pervasive passive monitoring [RFC7258].¶
A secondary (but significant) goal is to provide for ease of implementation, deployment and operation. This mechanism is expected to have a minimal impact upon performance, and require a trivial administrative effort to configure.¶
Preventing active attacks (such as a Man-in-the-Middle) is a non-goal for this specification. Furthermore, this specification is not intended to replace or offer an alternative to https, since https both prevents active attacks and invokes a more stringent security model in most clients.¶
An origin server that supports the resolution of http URIs can indicate support for this specification by providing an alternative service advertisement [RFC7838] for a protocol identifier that uses TLS, such as h2 [RFC7540]. Such a protocol MUST include an explicit indication of the scheme of the resource. This excludes HTTP/1.1; HTTP/1.1 clients are forbidden from including the absolute form of a URI in requests to origin servers (see Section 5.3.1 of [RFC7230]).¶
A client that receives such an advertisement MAY make future requests intended for the associated origin [RFC6454] to the identified service (as specified by [RFC7838]), provided that the alternative service opts in as described in Section 2.1.¶
A client that places the importance of protection against passive attacks over performance might choose to withhold requests until an encrypted connection is available. However, if such a connection cannot be successfully established, the client can resume its use of the cleartext connection.¶
A client can also explicitly probe for an alternative service advertisement by sending a request that bears little or no sensitive information, such as one with the OPTIONS method. Likewise, clients with existing alternative services information could make such a request before they expire, in order minimize the delays that might be incurred.¶
Client certificates are not meaningful for URLs with the http scheme, and therefore clients creating new TLS connections to alternative services for the purposes of this specification MUST NOT present them. A server that also provides https resources on the same port can request a certificate during the TLS handshake, but it MUST NOT abort the handshake if the client does not provide one.¶
It is possible that the server might become confused about whether requests’ URLs have a http or https scheme, for various reasons; see Section 4.4. To ensure that the alternative service has opted into serving http URLs over TLS, clients are required to perform additional checks before directing http requests to it.¶
Clients MUST NOT send http requests over a secured connection, unless the chosen alternative service presents a certificate that is valid for the origin as defined in [RFC2818]. Using an authenticated alternative service establishes “reasonable assurances” for the purposes of [RFC7838]. In addition to authenticating the server, the client MUST have obtained a valid http-opportunistic response for an origin (as per Section 2.3) using the authenticated connection. An exception to the latter restriction is made for requests for the “http-opportunistic” well-known URI.¶
For example, assuming the following request is made over a TLS connection that is successfully authenticated for those origins, the following request/response pair would allow requests for the origins “http://www.example.com” or “http://example.com” to be sent using a secured connection:¶
HEADERS + END_STREAM + END_HEADERS :method = GET :scheme = http :authority = example.com :path = /.well-known/http-opportunistic HEADERS :status = 200 content-type = application/json DATA + END_STREAM [ "http://www.example.com", "http://example.com" ]
Though this document describes multiple origins, this is only for operational convenience. Only a request made to an origin (over an authenticated connection) can be used to acquire this resource for that origin. Thus in the example, the request to http://example.com cannot be assumed to also provide an http-opportunistic response for http://www.example.com.¶
Clients MUST NOT send http requests and https requests on the same connection. Similarly, clients MUST NOT send http requests for multiple origins on the same connection.¶
This specification defines the “http-opportunistic” well-known URI [RFC5785]. A client is said to have a valid http-opportunistic response for a given origin when:¶
A client MAY treat an “http-opportunistic” resource as invalid if values it contains are not strings.¶
This document does not define semantics for “http-opportunistic” resources on an https origin, nor does it define semantics if the resource includes https origins.¶
Allowing clients to cache the http-opportunistic resource means that all alternative services need to be able to respond to requests for http resources. A client is permitted to use an alternative service without acquiring the http-opportunistic resource from that service.¶
A client MUST NOT use any cached copies of an http-opportunistic resource that was acquired (or revalidated) over an unauthenticated connection. To avoid potential errors, a client can request or revalidate the http-opportunistic resource before using any connection to an alternative service.¶
Clients that use cached http-opportunistic responses MUST ensure that their cache is cleared of any responses that were acquired over an unauthenticated connection. Revalidating an unauthenticated response using an authenticated connection does not ensure the integrity of the response.¶
User Agents MUST NOT provide any special security indicators when an http resource is acquired using TLS. In particular, indicators that might suggest the same level of security as https MUST NOT be used (e.g., a “lock device”).¶
A downgrade attack against the negotiation for TLS is possible.¶
For example, because the Alt-Svc header field [RFC7838] likely appears in an unauthenticated and unencrypted channel, it is subject to downgrade by network attackers. In its simplest form, an attacker that wants the connection to remain in the clear need only strip the Alt-Svc header field from responses.¶
Cached alternative services can be used to track clients over time; e.g., using a user-specific hostname. Clearing the cache reduces the ability of servers to track clients; therefore clients MUST clear cached alternative service information when clearing other origin-based state (i.e., cookies).¶
HTTP implementations and applications sometimes use ambient signals to determine if a request is for an https resource; for example, they might look for TLS on the stack, or a server port number of 443.¶
This might be due to expected limitations in the protocol (the most common HTTP/1.1 request form does not carry an explicit indication of the URI scheme and the resource might have been developed assuming HTTP/1.1), or it may be because how the server and application are implemented (often, they are two separate entities, with a variety of possible interfaces between them).¶
Any security decisions based upon this information could be misled by the deployment of this specification, because it violates the assumption that the use of TLS (or port 443) means that the client is accessing a HTTPS URI, and operating in the security context implied by HTTPS.¶
Therefore, server implementers and administrators need to carefully examine the use of such signals before deploying this specification.¶
This specification requires that a server send both an Alternative Service advertisement and host content in a well-known location to send HTTP requests over TLS. Servers SHOULD take suitable measures to ensure that the content of the well-known resource remains under their control. Likewise, because the Alt-Svc header field is used to describe policies across an entire origin, servers SHOULD NOT permit user content to set or modify the value of this header.¶
Mike Bishop contributed significant text to this document.¶
Thanks to Patrick McManus, Stefan Eissing, Eliot Lear, Stephen Farrell, Guy Podjarny, Stephen Ludin, Erik Nygren, Paul Hoffman, Adam Langley, Eric Rescorla, Julian Reschke, Kari Hurtta, and Richard Barnes for their feedback and suggestions.¶