|
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. Instead of trusting a certification authority to have made this association correctly, the user might instead trust the authoritative DNS server for the domain name to make that association. 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 July 12, 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] (Dierks, T. and E. Rescorla, “The Transport Layer Security (TLS) Protocol Version 1.2,” August 2008.) and DTLS [4347bis] (Rescorla, E. and N. Modadugu, “Datagram Transport Layer Security version 1.2,” July 2010.). 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] (Richardson, M., “A Method for Storing IPsec Keying Material in DNS,” March 2005.) and keys for SSH are covered in [RFC4255] (Schlyter, J. and W. Griffin, “Using DNS to Securely Publish Secure Shell (SSH) Key Fingerprints,” January 2006.).
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] (Cooper, D., Santesson, S., Farrell, S., Boeyen, S., Housley, R., and W. Polk, “Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile,” May 2008.) 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] (Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “DNS Security Introduction and Requirements,” March 2005.), [RFC4034] (Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “Resource Records for the DNS Security Extensions,” March 2005.), and [RFC4035] (Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “Protocol Modifications for the DNS Security Extensions,” March 2005.)), 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] (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.).
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 domain name for the TLSA RR can return one or more records of the type TLSA. The TLSA RRType is TBD.
The format of the data in the resource record is a binary record with three values, which MUST be in the order defined here:
1 -- Hash of an end-entity certificate
2 -- Full end-entity certificate in DER encoding
3 -- Hash of an certification authority's certificate
4 -- Full certification authority's certificate in DER encoding
Certificate types 1 through 4 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. [[ Later: maybe make yet-another-probably-never-used IANA registry for certificate types. ]]
The TLS client determines whether or not the certificate offered by the TLS server matches the certificate association 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.
For types 1 and 3, the hash used in the comparison is the hash type from the TLSA data.
Types 1 (hash of an end-entity certificate) and 2 (full end-entity certificate) are matched against the first certificate offered by the TLS server. For type 1, the certificate association is valid if the hash of the first certificate offered by the TLS server matches the value from the resource record. For type 2, the certificate association is valid if the certificate in the TLSA data matches to the first certificate offered by TLS.
Type 3 (hash of certification authority's certificate) can be used in one of two ways. If the hash of any certificate past the first in the certificate bundle from TLS matches the value from the TLSA data, and the chain in the certificate bundle is valid up to that certificate, then the certificate association is valid. Alternately, if the first certificate offered chains to a trust anchor, and the hash of that trust anchor matches the value from the TLSA data (assuming that the trust anchor is kept in certificate format), then the certificate association is valid.
Type 4 (full certification authority's certificate) 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 certificate from the TLSA data (assuming that the trust anchor is kept in certificate format).
[[ Need discussion of self-signed certificates being CA certificates. Need to be sure that this discussion uses correct PKIX terminology and is carefully explained. ]]
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 hash of an end-entity certificate:
www.example.com. IN TLSA (
1 2 e77b719d4e9c63c0b0a0333be0a4188e490b618e )
The use of mnemonics instead of numbers is not allowed.
[[ We could consider using Base64 instead of hex. ]]
[[ Need to do this, clearly. ]]
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 a match between the certificate association(s) and the server's end entity certificate in TLS is not found, the TLS client MUST abort the handshake with an "access_denied" error.
TLS client policy is deliberately not prescribed by this specification. A client MAY choose to trust a DNSSEC-secured certificate association, depending on its local policy.
[[ The preceding paragraph is probably wrong in the sense that it means that we now have no conformance requirements. There is probably no reason to even use this protocol unless you are going to fully trust the results. The one exception that has been discussed is that you might want to use the TLSA data as a "second positive opinion", such as in a GUI or in logging. Both of those seem fairly useless in the case of DNS resolution. Thus, the above paragraph may be changed by the WG in a future version of this draft. ]]
One expected use case for this protocol is that some TLS servers will begin to use self-signed certificates in association with certificate associations. A TLS client that is using this protocol needs to treat self-signed certificates as special, and thus SHOULD NOT attempt certificate validation on them. (An exception to this rule would be clients that keep self-signed end entity certificates in its trust anchor store.)
All data in a self-signed certificate other than the key itself can be ignored as untrusted unless a client validates the self-signed certificate to a trust anchor that is identical to the certificate. That means that the host name given in the self-signed certificate is meaningless, and that the only way to associate the public key in the certificate with the domain name is through the certificate association made in the DNS.
If a TLS client fully trusts the association between a domain name and the certificate that was provided by the DNS, then that client MUST ignore the domain name that is given in the self-signed certificate. That is, the certificate might contain a domain name that is different than the one that was used to get the TLSA data, but if the client is trusting the TLSA data, it doesn't matter what domain name is used in the certificate. An expected use case for this protocol is to allow someone who controls the private key on a certificate to use that certificate for multiple TLS servers. These servers might be on a single computer that has many domain names (such as a computer that is both a web host and a mail host, and is known by both "www.example.com" and "smtp.example.com"), or they might be on different computers (such as multiple computers that all respond IP addresses reachable as "www.example.com").
[[ Add more about virtual hosting and SNI TLS extension. ]]
Another expected use case for this protocol is that some TLS servers will use certificates that chain to a trust anchor that might not be one that is trusted by the TLS client, such as a local certification authority (CA) that is administered by the organization that runs the TLS server; this is a likely use for certificate types 3 and 4. Because of this, a TLS client that is using this protocol that performs certificate validation on server certificates MAY have a method to communicate with the user that differentiates between validation failures that occur on certificates that have had secure certificate associations and those that have not. If it does not have such a method of communication, the failure to validate SHOULD cause the same error as for any other certificate validation.
Some TLS clients extract data from the certificate other than the key to show to the user; for example, most modern web browsers have the ability to show an extended validation (EV) name that is embedded in a certificate. Because this data comes from a trusted third party and not the TLS server itself, TLS clients that extract additional information from TLS server certificates MUST validate those certificates in the normal fashion.
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.
The security of the protocols described in this document relies on the security of DNSSEC as used by the client requesting A and TLSA records.
A DNS administrator who goes rogue and changes both the A 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.
[[ Add discussion of the idea that TLSA makes things worse if an intermediate CA is compromised. Need more from Stephen Farrell. ]]
[[ Add discussion of length check to avoid potential issues with appended data. Need more from Carl Wallace. ]]
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, and Ondrej Sury.
| [4347bis] | Rescorla, E. and N. Modadugu, “Datagram Transport Layer Security version 1.2,” draft-ietf-tls-rfc4347-bis (work in progress), July 2010. |
| [RFC2119] | Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” BCP 14, RFC 2119, March 1997 (TXT, HTML, XML). |
| [RFC4033] | Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “DNS Security Introduction and Requirements,” RFC 4033, March 2005 (TXT). |
| [RFC4034] | Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “Resource Records for the DNS Security Extensions,” RFC 4034, March 2005 (TXT). |
| [RFC4035] | Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “Protocol Modifications for the DNS Security Extensions,” RFC 4035, March 2005 (TXT). |
| [RFC5246] | Dierks, T. and E. Rescorla, “The Transport Layer Security (TLS) Protocol Version 1.2,” RFC 5246, August 2008 (TXT). |
| [RFC5280] | Cooper, D., Santesson, S., Farrell, S., Boeyen, S., Housley, R., and W. Polk, “Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile,” RFC 5280, May 2008 (TXT). |
| [RFC4025] | Richardson, M., “A Method for Storing IPsec Keying Material in DNS,” RFC 4025, March 2005 (TXT). |
| [RFC4255] | Schlyter, J. and W. Griffin, “Using DNS to Securely Publish Secure Shell (SSH) Key Fingerprints,” RFC 4255, January 2006 (TXT). |
| Paul Hoffman | |
| VPN Consortium | |
| Email: | paul.hoffman@vpnc.org |
| Jakob Schlyter | |
| Kirei AB | |
| Email: | jakob@kirei.se |