| Network Working Group | P. Hoffman |
| Internet-Draft | VPN Consortium |
| Intended status: Standards Track | J. Schlyter |
| Expires: August 27, 2011 | Kirei AB |
| February 23, 2011 |
Using Secure DNS to Associate Certificates with Domain Names For TLS
draft-ietf-dane-protocol-05
TLS and DTLS use certificates for authenticating the server. Users want their applications to verify that the certificate provided by the TLS server is in fact associated with the domain name they expect. DNSSEC provides a mechanism for a zone operator to sign DNS information directly. This way, bindings of keys to domains are asserted not by external entities, but by the entities that operate the DNS. This document describes how to use secure DNS to associate the TLS server's certificate with the the intended domain name.
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 August 27, 2011.
Copyright (c) 2011 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.
The first response from the server in TLS may contain a certificate. In order for the TLS client to authenticate that it is talking to the expected TLS server, the client must validate that this certificate is associated with the domain name used by the client to get to the server. Currently, the client must extract the domain name from the certificate, must trust a trust anchor upon which the server's certificate is rooted, and must successfully validate the certificate.
Some people want a different way to authenticate the association of the server's certificate with the intended domain name without trusting the CA. Given that the DNS administrator for a domain name is authorized to give identifying information about the zone, it makes sense to allow that administrator to also make an authoritative binding between the domain name and a certificate that might be used by a host at that domain name. The easiest way to do this is to use the DNS.
This document applies to both TLS [RFC5246] and DTLS [4347bis]. In order to make the document more readable, it mostly only talks about "TLS", but in all cases, it means "TLS or DTLS". This document only relates to securely associating certificates for TLS and DTLS with host names; other security protocols are handled in other documents. For example, keys for IPsec are covered in [RFC4025] and keys for SSH are covered in [RFC4255].
In this document, a certificate association is based on a cryptographic hash of a certificate (sometimes called a "fingerprint") or on the certificate itself. For a fingerprint, a hash is taken of the binary, DER-encoded certificate, and that hash is the certificate association; the type of hash function used can be chosen by the DNS administrator. When using the certificate itself in the certificate association, the entire certificate in the normal format is used. This document also only applies to PKIX [RFC5280] certificates.
Certificate associations are made between a certificate or the hash of a certificate and a domain name. Server software that is running TLS that is found at that domain name would use a certificate that has a certificate association given in the DNS, as described in this document. A DNS query can return multiple certificate associations, such as in the case of different server software on a single host using different certificates (even if they are normally accessed with different host names), or in the case that a server is changing from one certificate to another.
This document defines a secure method to associate the certificate that is obtained from the TLS server with a domain name using DNS protected by DNSSEC. Because the certificate association was retrieved based on a DNS query, the domain name in the query is by definition associated with the certificate.
DNSSEC, which is defined in RFCs 4033, 4034, and 4035 ([RFC4033], [RFC4034], and [RFC4035]), uses cryptographic keys and digital signatures to provide authentication of DNS data. Information retrieved from the DNS and that is validated using DNSSEC is thereby proved to be the authoritative data. The DNSSEC signature MUST be validated on all responses in order to assure the proof of origin of the data.
This document only relates to securely getting the DNS information for the certificate association using DNSSEC; other secure DNS mechanisms are out of scope.
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 RFC 2119 [RFC2119].
A note on terminology: Some people have said that this protocol is a form of "certificate exclusion". This is true, but in a very unusual sense. That is, a DNS reply that contains two of the certificate types defined here inherently excludes every other possible certificate in the universe other than those found with a pre-image attack against one of those two. The certificate type defined here is better thought of as "enumeration" of a small number of certificate associations, not "exclusion" of a near-infinite number of other certificates.
Some of the terminology in this draft may not match with the terminology used in RFC 5280. This will be fixed in future versions of this draft, with help from the PKIX community. In specific, we need to say (in a PKIX-appropriate way) that when we say "valid up to" and "chains to", full RFC 5280 path processing including revocation status checking is intended.
This document defines a new DNS resource record type, "TLSA". A query on a prepared domain name for the TLSA RR can return one or more records of the type TLSA. The TLSA RRType is TBD.
Domain names are prepared for requests in the following manner.
For example, to request a TLSA resource record for an HTTP server running TLS on port 443 at "www.example.com", you would use "_443._tcp.www.example.com" in the request. To request a TLSA resource record for an SMTP server running the STARTTLS protocol on port 25 at "mail.example.com", you would use "_25._tcp.mail.example.com".
The format of the data in the resource record is a binary record with three values, which MUST be in the order defined here:
Using the same hash algorithm as is used in the signature in the certificate will make it more likely that the TLS client will understand this TLSA data.
Certificate types 1 and 2 explicitly only apply to PKIX-formatted certificates. If TLS allows other formats later, or if extensions to this protocol are made that accept other formats for certificates, those certificates will need certificate types.
A TLS client conforming to this protocol MUST treat the certificate for association in a TLSA resource record for a domain name as a trust anchor for that domain name at the specific port number and transport name that was queried. This trust anchor MUST only be used for the domain name of the resource record. The trust anchor MUST NOT be loaded for longer than the TTL on the TSLA record.
The TLS client determines whether or not the certificate offered by the TLS server matches the trust anchor received in the TLSA resource record. If the certificate from the TLS server matches, the TLS client accepts the certificate association. Each certificate type has a different method for determining matching.
Certificate type 1 (end-entity certificate) is matched against the first certificate offered by the TLS server. The certificate for association is used only for exact matching, not for chained validation. With reference type 0, the certificate association is valid if the certificate in the TLSA data matches to the first certificate offered by TLS. With reference types other than 0, the certificate association is valid if the hash of the first certificate offered by the TLS server matches the value from the TLSA data.
Certificate type 2 (certification authority's certificate) can be used in one of two ways. With reference type 0, the certificate in the TLSA resource record is used in chaining from the end-entity given in TLS. The certificate association is valid if the first certificate in the certificate bundle can be validly chained to the trust anchor from the TLSA data. With reference types other than 0, if the hash of any certificate past the first in the certificate bundle from TLS matches the trust anchor from the TLSA data, and the chain in the certificate bundle is valid up to that TLSA trust anchor, then the certificate association is valid. Alternately, if the first certificate offered chains to an existing trust anchor in the TLS client's trust anchor repository, and the hash of that trust anchor matches the value from the TLSA data, then the certificate association is valid.
[[ Need discussion of self-signed certificates being CA certificates. Need to be sure that this discussion uses correct PKIX terminology and is carefully explained. ]]
_443._tcp.www.example.com. IN TLSA (
1 2 5c1502a6549c423be0a0aa9d9a16904de5ef0f5c98
c735fcca79f09230aa7141 )
The RDATA of the presentation format of the TLSA resource record consists of two numbers (certificate and hash type) followed by the bytes containing the certificate or the hash of the associated certificate itself, presented in hex. An example of a SHA-256 hash (type 2) of an end-entity certificate (type 1) would be:
_443._tcp.www.example.com. IN TLSA ( 2 0 308202c5308201ada00302010202090... )
An example of an unhashed CA certificate (type 2) would be:
Because the length of hashes and certificates can be quite long, presentation format explicitly allows line breaks and white space in the hex values; those characters are removed when converting to the wire format.
1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Cert type | Hash type | /
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ /
/ /
/ Certificate for association /
/ /
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The wire format is:
_443._tcp.www.example.com. IN TYPE65534 \# 34 ( 01025c1502a6549c42
3be0a0aa9d9a16904de5ef0f5c98c735fcca79f09230aa7141 )
The wire format for the RDATA in the first example given above would be:
_443._tcp.www.example.com. IN TYPE65534 \# 715 0200308202c5308201a...
The wire format for the RDATA in the second example given above would be:
Note that in the preceding examples, "TYPE65534" is given as an example. That RR Type is in the IANA "private use" range; the real RR Type for TLSA will be issued by IANA, as described in the IANA Considerations section below.
In order to use one or more TLS certificate associations described in this document obtained from the DNS, an application MUST assure that the certificates were obtained using DNS protected by DNSSEC. TLSA records must only be trusted if they were obtained from a trusted source. This could be a localhost DNS resolver answer with the AD bit set, an inline validating resolver library primed with the proper trust anchors, or obtained from a remote nameserver to which one has a secured channel of communication.
If a certificate association contains a hash type that is not understood by the TLS client, that certificate association MUST be marked as unusable.
An application that requests TLS certificate associations using the method described in this document obtains zero or more usable certificate associations. If the application receives zero usable certificate associations, it processes TLS in the normal fashion.
If a match between one of the certificate association(s) and the server's end entity certificate in TLS is found, the TLS client continues the TLS handshake. If no match between the usable certificate association(s) and the server's end entity certificate in TLS is found, the TLS client MUST abort the handshake with an "access_denied" error.
This document uses a new DNS RRType, TLSA, whose value is TBD. A separate request for the RRType will be submitted to the expert reviewer, and future versions of this document will have that value instead of TBD.
Value Short description Ref. ------------------------------------------------------------- 0 Reserved [This] 1 End-entity certificate [This] 2 CA's certificate [This] 3-254 Unassigned 255 Private use
This document creates a new registry, "Certificate Types for TLSA Resource Records". The registry policy is "RFC Required". The initial entries in the registry are:
Applications to the registry can request specific values that have yet to be assigned.
Value Short description Ref. ----------------------------------------------------- 0 No hash used [This] 1 SHA-1 NIST FIPS 180-2 2 SHA-256 NIST FIPS 180-2 3 SHA-384 NIST FIPS 180-2 4-254 Unassigned 255 Private use
This document creates a new registry, "Hash Types for TLSA Resource Records". The registry policy is "Specification Required". The initial entries in the registry are:
Applications to the registry can request specific values that have yet to be assigned.
The security of the protocols described in this document relies on the security of DNSSEC as used by the client requesting A/AAAA and TLSA records.
A DNS administrator who goes rogue and changes both the A/AAAA and TLSA records for a domain name can cause the user to go to an unauthorized server that will appear authorized, unless the client performs certificate validation and rejects the certificate. That administrator could probably get a certificate issued anyway, so this is not an additional threat.
The values in the TLSA data will be normally entered in the DNS through the same system used to enter A/AAAA records, and other DNS information for the host name. If the authentication for changes to the host information is weak, an attacker can easily change any of this information. Given that the TLSA data is not easily human-readable, an attacker might change those records and A/AAAA records and not have the change be noticed if changes to a zone are only monitored visually.
If the authentication mechanism for adding or changing TLSA data in a zone is weaker than the authentication mechanism for changing the A/AAAA records, a man-in-the-middle who can redirect traffic to their site may be able to impersonate the attacked host in TLS if they can use the weaker authentication mechanism. A better design for authenticating DNS would be to have the same level of authentication used for all DNS additions and changes for a particular host.
SSL proxies can sometimes act as a man-in-the-middle for TLS clients. In these scenarios, the clients add a new trust anchor whose private key is kept on the SSL proxy; the proxy intercepts TLS requests, creates a new TLS session with the intended host, and sets up a TLS session with the client using a certificate that chains to the trust anchor installed in the client by the proxy. In such environments, the TLSA protocol will prevent the SSL proxy from functioning as expected because the TLS client will get a certificate association from the DNS that will not match the certificate that the SSL proxy uses with the client. The client, seeing the proxy's new certificate for the supposed destination will not set up a TLS session.
Many of the ideas in this document have been discussed over many years. More recently, the ideas have been discussed by the authors and others in a more focused fashion. In particular, some of the ideas here originated with Paul Vixie, Dan Kaminsky, Jeff Hodges, Phill Hallam-Baker, Simon Josefsson, Warren Kumari, Adam Langley, Ilari Liusvaara, Scott Schmit, and Ondrej Sury.
| [RFC4025] | Richardson, M., "A Method for Storing IPsec Keying Material in DNS", RFC 4025, March 2005. |
| [RFC4255] | Schlyter, J. and W. Griffin, "Using DNS to Securely Publish Secure Shell (SSH) Key Fingerprints", RFC 4255, January 2006. |