| Using TLS in Applications | S.F. Friedl |
| Internet-Draft | Cisco Systems, Inc. |
| Intended status: Standards Track | T.K. Kaupe |
| Expires: September 14, 2014 | Microsoft Corp. |
| S.G. Gorti | |
| Cisco Systems, Inc. | |
| March 13, 2014 |
TLS Certificate Identity Verification Procedure for SMTP MTA to MTA Connections
draft-friedl-uta-smtp-mta-certs-00
This document describes TLS server identity verification procedure for Message Transfer Agent (MTA) to Message Transfer Agent connections in an SMTP email network, with specific guidance on identity verification steps associated with delegated email services. This document is intended to supplement the identify verification procedures described in [RFC6125].
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 14, 2014.
Copyright (c) 2014 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 TLS security protocol [RFC5246] is typically used to encapsulate application protocols to provide privacy, data integrity and server identity verification between two application servers which would otherwise be conducting their communication in plain text over potentially insecure network links. There are two layers:
There are various RFCs, such as [RFC6125] which describe TLS in the context of more general domain based applications, with the canonical example being a user-agent (e.g. browser) securely accessing content on a server (e.g. web server) [RFC2818]. RFCs also describe TLS in the more specific context of SMTP, however these RFCs focus on a user-agent (e.g. mobile or PC mail client) submitting email to a mail server using the STARTTLS SMTP extension[RFC2595] [RFC3207]. No RFCs explicitly describe TLS in the context of SMTP communication between two MTAs in an email network. There exist sufficient differences in the MTA-to-MTA scenario, particularly with respect to server identification, which warrant an explicit set of recommendations. This document discusses various strategies that a sending MTA can use for the identification and authentication of a destination MTA when transferring a message within an email network. It should be noted that an email network could be defined with MTA to Message Delivery Agent (MDA) connections, in which case the same verification and authentication rules should apply to the MTA to MDA scenario.
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 [RFC2119].
Two fundamental aspects govern how an MTA validates the identity of an email server when establishing a TLS session:
The destination MTA identity is verified through a process of ordered comparison of reference and presented identity pairs in conformance to the rules defined in [RFC6125].
In the case of an MTA, the reference identity of the destination email server is typically expressed via the MX record for the recipient email domain. A user addresses email to a recipient at a specific email domain (e.g. recipient@example.net) and the MTA then performs a DNS MX query using the [RFC5321] domain section of the RCPT TO command forward path to determine the email server hostname (e.g. mail.example.net) for the recipient email domain. It is important to note that the email server name will not necessarily be a subdomain of the recipient email domain; this will never be the case with delegated email hosting.
The server presented identity SHOULD be a SubjectAltName (SAN) of type DNSName of a X.509 public key certificate. In keeping with [RFC6125], SAN entries SHOULD be used for the presented identity, but the CN entry of a X.509 public key certificate MAY be used for backwards compatibility with deployed infrastructure if no SAN entries exist in the certificate.
[RFC6125] recommends deprecating support for SAN entries that include wildcards for two primary reasons: (1) the lack of clarity in existing specification as to the allowable locations of wildcard characters and (2) the fact that a wildcard SAN entries vouches for all servers in a domain, including possibly rogue or buggy servers. In the case of MTA to delegated email service connections, this document proposes continued support for DNSNames containing wildcards as wildcard DNSNames are needed to support some delegated email hosting scenarios.
For delegated hosting, some third parties may use the same email server hostname(s) for all the domains that they host:
Alternatively, a third party email service might use a unique hostname for each email domain:
If unique hostnames are associated with each email domain, then there will be as many host names as email domains and it will not be possible to include all hostnames as SAN DNSName entries in a certificate. Wildcarded SAN entries would then be the only mechanism by which a single certificate may be used for all email server hostname reference identities. If all the email server hostnames are of the form [domainidentifier].[hostnameroot] (e.g. example1net.mail.delegatedexample.net), then the SAN SHOULD include a DNSName entry of the form *.[hostnameroot] (e.g. *.mail.delegatedexample.net). The email server certificate SHOULD NOT include a wildcard character in the presented identity in any position other than the left-most label.
Validation of the server identity entails a comparison of the reference identity to the identity presented in the server certificate. There are several approaches for validation, given the various reference identifiers that may be used and the fundamental difference between the self-hosted and delegated hosting models.
The reference identity for an email server SHOULD be determined using the following methods:
An MTA SHOULD validate an email server identity when an exact match exists between the presented identity as a SAN DNSName entry in the email server's certificate and the email server's reference identity.
An MTA SHOULD validate an email server identity when a [RFC6125] Section 6.4.3 compliant wildcard match exists between the presented identity as a SAN DNSName entry with wildcards in the email server's certificate and the email server's reference identity.
For backward compatibility, if no SAN DNSName entries exist in an email server's certificate but a CN entry exists then an MTA SHOULD validate an email server identity if an exact match exists between the CN and the email server's reference identity.
For backward compatibility, if no SAN DNSName entries exist in an email server's certificate but a CN entry exists then an MTA SHOULD validate an email server identity when a [RFC6125] Section 6.4.3 compliant wildcard match exists between the presented identity as a CN entry with wildcards in the email server's certificate and the email server's reference identity.
The order of evaluation of the different methods for an MTA to validate and email server identity are important and an MTA SHOULD use the following ordering of matching tests:
If one or more SAN DNSName entries are present in the email server's certificate the following matching tests SHOULD be used in the order specified:
If one or more SAN DNSName entries are present in the email server's certificate and none of the matching tests specified above pass, then the MTA will have failed to validate the email server's identity. and the MTA SHOULD log an error indicating that the validation process failed.
An MTA MUST NOT validate an email server identity against the CN entry of an email server's certificate if there exist one or more SAN DNSName entries in the certificate. If and only if no SAN DNSName entries exist in the email server certificate and a CN is present in the email server's certificate the same matching process detailed in 4.1.4.6.1 above MUST be used with the CN as the presented identity instead of the SAN.
This document addresses only the procedure by which an MTA should verify the identity of an email server with which it wishes to establish a TLS connection. The procedures described in this document do nothing to address or ameliorate the fundamental security risk associated with obtaining an email server name through an insecure MX query. In the abscence of DNSSEC there exist a large number of techniques whereby a false email server name could be returned to the MTA through an insecure MX query. It should be noted that in the absence of DNSSEC, the MX query is no more risk prone than a browser A lookup. Use of TLS eliminates risks of passive listening and imposes a requirement that an attacker to obtain a certificate that will be trusted by the sending MTA and actively participate in the session by terminating the TLS session requested by the sending MTA.
Risks associated with insecure DNS MX lookups may be ameliorated by explicit association of the email server name to an email domain in the MTA configuration.
This document includes no request to IANA.
Thanks to Dan Wing for multiple reviews of this draft and valuable suggestions for improving it.