| Internet-Draft | Key Consistency Double Check | October 2022 |
| Schwartz | Expires 22 April 2023 | [Page] |
Several recent IETF privacy protocols require clients to acquire bootstrap information for a service in a way that guarantees both authenticity and consistency, e.g., encrypting to the same key as many other users. This specification defines a procedure for transferring arbitrary HTTP resources in a manner that provides these guarantees. The procedure relies on access to a semi-trusted HTTP proxy, under the same security assumptions as an Oblivious HTTP Relay.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-schwartz-ohai-consistency-doublecheck/.¶
Source for this draft and an issue tracker can be found at https://github.com/bemasc/access-services.¶
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 22 April 2023.¶
Copyright (c) 2022 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.¶
Oblivious HTTP [I-D.ietf-ohai-ohttp] enables an HTTP client to send requests to a target in such a way that no single party can determine both the identity of the client and the contents of its request. Oblivious HTTP identifies four parties to each exchange: the Client, Relay, Gateway and Target.¶
In order for Oblivious HTTP's privacy guarantees to hold, several preconditions must be met:¶
The Privacy Pass protocol [I-D.ietf-privacypass-protocol] allows a Client to retrieve tokens from an Issuer, and present them to an Origin, in such a way that the Origin can verify the validity of the tokens but cannot identify the Client, even if it colludes with the Issuer. Privacy Pass requires a similar set of preconditions for its privacy guarantees:¶
Pre-requisite 1 on this list can be achieved by employing a shared transport proxy, on the assumption that at least one of the proxy or the Origin is non-malicious. (This is the same assumption that an Oblivious HTTP Client makes about its Relay).¶
This specification assumes that a shared, "semi-trusted" proxy is available (fulfilling precondition 1 on each list). It defines a three-party protocol for a Client, Proxy, and Origin to fetch an HTTP resource in a manner that ensures authenticity and consistency among Clients of a single Proxy. When used to initialize Oblivious HTTP or Privacy Pass, it guarantees preconditions 2, 3, and optionally 4.¶
The input to this protocol is a Desired Resource URI: an "https" URI on the Origin that is assumed to have been distributed to Clients in a globally consistent fashion. For example, this URI might be the default value of a software setting, or it might be published on a third party's website. This specification allows Clients to convert the static, long-lived Desired Resource URI into a fresh copy of that resource, i.e. to fetch the resource.¶
In principle, the Desired Resource could have been distributed directly through the assumed globally consistent channel. However, these ad hoc publication channels may not be fast enough to support frequent updates (e.g., key rotations), especially if updates require user intervention.¶
This draft is an instantiation of the Shared Proxy with Key Confirmation strategy defined in Section 4.3 of [I-D.wood-key-consistency].¶
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.¶
In the Key Consistency Double-Check procedure, the Client emits two HTTP GET requests for the Desired Resource. One uses the Proxy as an HTTP request proxy (terminating TLS and HTTP), and the other optionally uses it as a transport proxy (using TLS end-to-end with the Origin). The Proxy will forward the first request to the Origin only if the response is not in cache.¶
+--------+ +-------+ +--------+
| |<=====>| |<----->| |
| Client | | Proxy | | Origin |
| |<=====================>| |
+--------+ +-------+ +--------+
The Proxy caches the response, ensuring that all clients share it during its freshness lifetime. The client checks this against the authenticated response from the Origin, preventing forgeries.¶
The Desired Resource MUST include a "strong validator" ETag (Section 2 of [RFC7232]) in any response to a GET request, and MUST support the "If-Match" HTTP request header (Section 3 of [RFC7232]). The response MUST indicate "Cache-Control: public, no-transform, s-maxage=(...), immutable" [RFC9111][RFC8246]. For efficiency reasons, the max age SHOULD be at least 60 seconds, and preferably much longer.¶
If the Desired Resource changes, and the Origin receives a request for the resource whose "If-Match" header identifies a previously served version that has not yet expired, it MUST return a success response containing the previous version. This response MAY indicate "Cache-Control: private".¶
The Proxy MUST offer HTTP request proxy capability, and SHOULD also offer TCP proxy, UDP proxy, and encrypted DNS resolution capabilities. (An associated encrypted DNS resolver enables reliable use of HTTPS records [SVCB], improves metadata confidentiality, and allows EDNS Client Subnet to be disabled reliably.) For example, a proxy in the recommended configuration could be described by the following Access Service Description [I-D.schwartz-masque-access-descriptions]:¶
{
"http": {
"template": "https://relay.example.org/http{?target_uri}"
},
"tcp": {
"template":
"https://proxy.example.org/tcp{?target_host,tcp_port}"
},
"udp": {
"template":
"https://proxy.example.org/masque{?target_host,target_port}"
},
"dns": {
"template": "https://doh.example.com/dns-query{?dns}",
}
}
The Proxy MUST allow use of the GET method to proxy small responses, and SHOULD make ample cache space available in order to avoid eviction of Desired Resources. The proxy SHOULD share cache state among all clients, to ensure that they observe the same resource contents. If the cache must be partitioned for architectural or performance reasons, operators SHOULD keep the number of users in each partition as large as possible.¶
Proxies MUST preserve the ETag response header on cached responses, and MUST add an Age header ([RFC9111], Section 5.1) to all proxied responses. Proxies MUST respect the "Cache-Control: immutable" directive, and MUST NOT revalidate fresh immutable cache entries in response to any incoming requests. (Note that this is different from the general recommendation in Section 2.1 of [RFC8246]). Proxies also MUST NOT accept PUSH_PROMISE frames from the target.¶
Proxies SHOULD employ defenses against malicious attempts to fill the cache. Some possible defenses include:¶
Proxies that are not intended for general-purpose use MAY impose strict transfer limits or rate limits on transport proxy usage.¶
If the Proxy offers an Encrypted DNS service, it MUST NOT enable EDNS Client Subnet support [RFC7871].¶
The Client is assumed to know the "https" URI of the Desired Resource. To retrieve that resource, it MUST perform the following "double-check" procedure:¶
This procedure ensures that the retrieved copy of the Desired Resource is authentic and is identical to the one held by any other users of this proxy. Once response A or B expires, the client MUST refresh it before continuing to use this resource, and MUST repeat the "double-check" process if either response changes.¶
Clients MUST perform each fetch to the Origin (step 4) as a fully isolated request. Any state related to this Origin (e.g., cached DNS records, CONNECT-UDP tunnels, QUIC transport state, TLS session tickets, HTTP cookies) MUST NOT be shared with prior or subsequent requests.¶
If the Desired Resource does not exist, the Double Check requirements apply without modification. The Cache-Control headers ensure that the negative response (e.g., HTTP 404) is cacheable regardless of its status code. If Double Check succeeds with a negative response, the Client can be confident that any other Clients of this proxy are holding the same negative response.¶
In this example, the client has been configured with a Proxy via the following Access Service Description:¶
{
"dns": {
"template": "https://proxy.example.org/dns-query{?dns}",
},
"udp": {
"template":
"https://proxy.example.org/masque{?target_host,target_port}"
},
"http": {
"template": "https://relay.example.org/http{?target_uri}"
}
}
¶
The client has recently been instructed to use an Encrypted DNS server identified as "doh.example.com" that might support Oblivious DoH. To discover and access this Oblivious DoH service, the client attempts to retrieve two Desired Resources on the Origin "https://doh.example.com":¶
The DNS Server's Access Service Description, at "/.well-known/access-services" (Section 5 of [I-D.schwartz-masque-access-descriptions]).¶
The Oblivious Gateway KeyConfig, at "/.well-known/oblivious-gateway" (Section 5 of [I-D.pauly-ohai-svcb-config]).¶
To prevent client-targeting attacks, the client retrieves both of these resources via DoubleCheck. The HTTP requests for the Access Service Description appears as follows:¶
HEADERS
:method = GET
:scheme = https
:authority = relay.example.org
:path = /http?target_uri=
https%3A%2F%2Fdoh.example.com%2F.well-known%2Faccess-services
accept: application/access-services+json
HEADERS
:status = 200
cache-control: public, immutable, \
no-transform, s-maxage=86400
age: 80000
etag: ABCD1234
content-type: application/access-services+json
{
"dns": {
"template":
"https://doh.example.com/foo{?dns}"
}
}
HEADERS
:method = CONNECT
:protocol = connect-udp
:scheme = https
:authority = proxy.example.org
:path = /masque?target_host=doh.example.com,target_port=443
capsule-protocol = ?1
HEADERS
:status = 200
capsule-protocol = ?1
¶
The client now has a CONNECT-UDP tunnel to doh.example.com, over which it performs the following GET request using HTTP/3:¶
HEADERS
:method = GET
:scheme = https
:authority = doh.example.com
:path = /.well-known/access-services
accept: application/access-services+json
if-match = ABCD1234
HEADERS
:status = 200
cache-control: public, immutable, \
no-transform, s-maxage=86400
etag: ABCD1234
content-type: application/access-services+json
{
"dns": {
"template":
"https://doh.example.com/foo{?dns}"
}
}
¶
Having successfully fetched the DoH Service Description from both locations, the client confirms that:¶
This concludes the DoubleCheck procedure. The Oblivious Gateway KeyConfig is retrieved by a similar procedure. These two DoubleCheck procedures can run in parallel, and MAY share a single transport proxy tunnel for efficiency.¶
Once the client has double-checked the DoH Service Description and the KeyConfig, it can use the Oblivious DoH service by forming DNS-over-HTTPS requests for "https://doh.example.com/foo{?dns}" as Binary HTTP requests, encrypting them to the KeyConfig, and POSTing the ciphertext to "https://relay.example.org/http?target_uri=https%3A%2F%2Fdoh.example.com%2F.well-known%2Foblivious-gateway".¶
Suppose that the Client-Proxy Round-Trip Time (RTT) is A, and the Proxy-Origin RTT is B. Suppose additionally that the Client has a persistent connection to the Proxy that is already running. Then the procedure described in Section 4.3, with a CONNECT-UDP transport proxy, requires:¶
A for the GET request to the Proxy¶
A for the CONNECT-UDP request to the Proxy¶
A + B for a QUIC handshake to the Origin¶
A + B for the GET request to the Origin¶
This is a total of 4A + 4B in the worst case. However, clients can reduce the latency by issuing the requests to the Proxy in parallel, and by using CONNECT-UDP's "false start" support. The Origin can also optimize performance, by issuing long-lived TLS session tickets. With these optimizations, the expected total time is 2A + 2B.¶
This procedure only needs to be repeated if the Desired Resource has expired. To enable regular key rotation and operational adjustments, a cache lifetime of 24 hours may be suitable. Clients MAY perform this procedure in advance of an expiration to avoid a delay.¶
All clients of the same Proxy and Desired Resource will have locally cached copies with the same expiration time. When this copy expires, all active clients will send refresh GET requests to the Proxy at their next request. Proxies SHOULD use "request coalescing" to avoid duplicate cache-refresh requests to the Origin.¶
If the Desired Resource has changed, these clients will all initiate GET requests to the Origin (via transport proxy if applicable) to double-check the new contents. Proxies and Origins MAY use an HTTP 503 response with a "Retry-After" header to manage load spikes.¶
A malicious Proxy could attempt to forge the Desired Resource by replacing the response body (e.g., returning an Oblivious HTTP KeyConfig containing a Public Key controlled by the Relay). This is prevented by the Client's requirement that the Desired Resource be served to it by the Origin over HTTPS (Section 4.3).¶
A malicious Origin could attempt to break the consistency guarantee by issuing each Client a unique, persistent variant of the Desired Resource. This attack is prevented by the Client's requirement that the resource be fresh according to the Proxy's cache (Section 4.3).¶
A malicious Origin could attempt to rotate its entry in the Proxy's cache in several ways:¶
By also acting as a Client and sending requests designed to replace the Desired Resource in the cache before it expires:¶
A malicious Origin could attempt to identify or link different requests for the Desired Resource, in order to link proxied requests that follow shortly after. If a transport proxy is used (as recommended), this is prevented by fully isolating each request (Section 4.3), and by disabling EDNS Client Subnet (Section 4.2).¶
A malicious Client could use the Proxy to send abusive traffic to any destination on the internet. Abuse concerns can be mitigated by imposing a rate limit at the proxy (Section 4.2).¶
This specification assumes that the Client starts with identities of the Proxy and the Desired Resource that are authentic and widely shared. If these identities are inauthentic, or are unique to the client, then the security goals of this specification are not achieved.¶
This specification assumes that at most a small fraction of Clients are acting on behalf of a malicious Origin. If a large fraction of the Clients are malicious, they could conspire to flood the Proxy's cache with entries that seem popular, leading to rapid eviction of any cached resources. Similar concerns apply if a malicious Origin can compel naive Clients to fetch a very large number of distinct resources through the Proxy.¶
Even when a transport proxy is used, the Client's requests for the Desired Resource may become linkable if they have distinctive TLS ClientHellos, QUIC Initials, HTTP/3 Settings, RTT, or other protocol features observable through the transport proxy. This specification does not offer specific mitigations for protocol fingerprinting.¶
No IANA action is requested.¶
TODO acknowledge.¶