The Signature HTTP Authentication Scheme

Existing HTTP authentication schemes are probeable in the sense that it is possible for an unauthenticated client to probe whether an origin serves resources that require authentication. It is possible for an origin to hide the fact that it requires authentication by not generating Unauthorized status codes, however that only works with non-cryptographic authentication schemes: cryptographic signatures require a fresh nonce to be signed, and there is no existing way for the origin to share such a nonce without exposing the fact that it serves resources that require authentication. This document proposes a new non-probeable cryptographic authentication scheme.¶

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 December 30, 2023.¶

Copyright (c) 2023 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶

HTTP authentication schemes (see Section 11 of [HTTP]) allow origins to restrict access for some resources to only authenticated requests. While these schemes commonly involve a challenge where the origin asks the client to provide authentication information, it is possible for clients to send such information unprompted. This is particularly useful in cases where an origin wants to offer a service or capability only to "those who know" while all others are given no indication the service or capability exists. Such designs rely on an externally-defined mechanism by which keys are distributed. For example, a company might offer remote employee access to company services directly via its website using their employee credentials, or offer access to limited special capabilities for specific employees, while making discovering (probing for) such capabilities difficult. Members of less well-defined communities might use more ephemeral keys to acquire access to geography- or capability-specific resources, as issued by an entity whose user base is larger than the available resources can support (by having that entity metering the availability of keys temporally or geographically).¶

While digital-signature-based HTTP authentication schemes already exist ([HOBA]), they rely on the origin explicitly sending a fresh challenge to the client, to ensure that the signature input is fresh. That makes the origin probeable as it send the challenge to unauthenticated clients. This document defines a new signature-based authentication scheme that is not probeable.¶

This document defines the "Signature" HTTP authentication scheme. It uses asymmetric cryptography. User agents possess a key ID and a public/private key pair, and origin servers maintain a mapping of authorized key IDs to their associated public keys.¶

The client uses a TLS keying material exporter to generate data to be signed (see Section 4) then sends the signature using the Authorization or Proxy-Authorization header field. The signature and additional information are exchanged using authentication parameters (see Section 5).¶

This authentication scheme is only defined for uses of HTTP with TLS [TLS]. This includes any use of HTTP over TLS as typically used for HTTP/2 [HTTP/2], or HTTP/3 [HTTP/3] where the transport protocol uses TLS as its authentication and key exchange mechanism [QUIC-TLS].¶

• The TLS version in use is greater or equal to 1.3 [TLS].
• The TLS version in use is 1.2 and the Extended Master Secret extension [RFC7627] has been negotiated.

Clients MUST NOT use the Signature authentication scheme on connections that do not meet one of the two properties above. If a server receives a request that uses this authentication scheme on a connection that meets neither of the above properties, the server MUST treat the request as malformed.¶

• the label is set to "EXPORTER-HTTP-Signature-Authentication"
• the context is set to the structure described in Section 4.1
• the exporter output length is set to 48 bytes (see Section 4.2)

This specification defines the following authentication parameters. These parameters use structured fields ([STRUCTURED-FIELDS]) in their definition, even though the Authorization field itself does not use structured fields. Due to the syntax requirements for authentication parameters, the byte sequences defined below SHALL be enclosed in double-quotes (the base64-encoded data and colon delimeters are enclosed in double-quotes, see example in Section 6).¶

• absence of the Authorization (or Proxy-Authorization) field
• failure to parse that field
• use of the Signature authentication scheme with an unknown key ID
• failure to validate the verification parameter
• failure to validate the signature.

In order to validate the signature, the server needs to first parse the field containing the signature, then look up the key ID in its database of public keys, and finally perform the cryptographic validation. These steps can take time, and an attacker could detect use of this mechanism if that time is observable by comparing the timing of a request for a known non-existent resource to the timing of a request for a potentially authenticated resource. Servers can mitigate this observability by slightly delaying responses to some non-existent resources such that the timing of the authentication verification is not observable. This delay needs to be carefully considered to avoid having the delay itself leak the fact that this origin uses this mechanism at all.¶

Non-probeable resources also need to be non-discoverable for unauthenticated users. For example, if a server operator wishes to hide an authenticated resource by pretending it does not exist to unauthenticated users, then the server operator needs to ensure there are no unauthenticated pages with links to that resource, and no other out-of-band ways for unauthenticated users to discover this resource.¶

• The intermediary can validate the authentication received from the client, then inform the upstream HTTP server of the presence of valid authentication.
• The intermediary can export the Signature Input and Verification (see Section 4.2}), and forward it to the upstream HTTP server, then the upstream server performs the validation.

The mechanism for the intermediary to communicate this information to the upstream HTTP server is out of scope for this document.¶

Note that both of these mechanisms require the upstream HTTP server to trust the intermediary. This is usually the case because the intermediary already needs access to the TLS certificate private key in order to respond to requests.¶

The Signature HTTP authentication scheme allows a user agent to authenticate to an origin server while guaranteeing freshness and without the need for the server to transmit a nonce to the user agent. This allows the server to accept authenticated clients without revealing that it supports or expects authentication for some resources. It also allows authentication without the user agent leaking the presence of authentication to observers due to clear-text TLS Client Hello extensions.¶

The authentication proofs described in this document are not bound to individual HTTP requests; if the key is used for authentication proofs on multiple requests on the same connection, they will all be identical. This allows for better compression when sending over the wire, but implies that client implementations that multiplex different security contexts over a single HTTP connection need to ensure that those contexts cannot read each other's header fields. Otherwise, one context would be able to replay the Authorization header field of another. This constraint is met by modern Web browsers. If an attacker were to compromise the browser such that it could access another context's memory, the attacker might also be able to access the corresponding key, so binding authentication to requests would not provide much benefit in practice.¶

Key material used for the Signature HTTP authentication scheme MUST NOT be reused in other protocols. Doing so can undermine the security guarantees of the authentication.¶

Origins offering this scheme can link requests that use the same key. However, requests are not linkable across origins if the keys used are specific to the individual origins using this scheme.¶

The authors would like to thank many members of the IETF community, as this document is the fruit of many hallway conversations. In particular, the authors would like to thank Nick Harper, Dennis Jackson, Ilari Liusvaara, Lucas Pardue, Justin Richer, Ben Schwartz, Martin Thomson, and Chris Wood for their reviews and contributions. The mechanism described in this document was originally part of the first iteration of MASQUE [MASQUE-ORIGINAL].¶