Secure inter-domain SMTP with TLS, DNSSEC and TLSA records.
draft-fanf-dane-smtp-00

Abstract

SMTP supports STARTTLS for inter-domain mail transfer, but it only provides very limited security because the server's certificate cannot be authenticated. This memo specifies how TLSA records in the DNS can be used for proper MX target server authentication.

Status of This Memo

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 October 31, 2012.

Copyright Notice

Copyright (c) 2012 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.

Table of Contents

• 1. Introduction
• 2. Terminology
• 3. Details of SMTP with TLSA
• 4. IANA Considerations
• 5. Security considerations
• 6. References
• 6.1. Normative References
• 6.2. Informative References
• Appendix A. Rationale - choice of certificate identity
• Author's Address

1. Introduction

The specification for SMTP over TLS [RFC3207] does not describe how to authenticate a server: which identity relating to the connection ought to be authenticated by the server's certificate. In practice, most certificates presented by publicly-referenced SMTP servers either cannot be validated with respect to a well-known certification authority, or do not verify any identity expected by the client.

As a result, inter-domain SMTP clients cannot require working server authentication if they want to successfully send mail using TLS. Therefore TLS currently provides only a limited amount of additional security for inter-domain SMTP. Its encryption protects against on-path passive eavesdropping; but it does not protect against an active attack, since the client has no way to detect when an attacker is spoofing the server.

This memo describes how to fix this using DNSSEC [RFC4033] and TLSA records [I-D.ietf-dane-protocol].

We use DNSSEC to secure the association between a mail domain and its SMTP server host names. A server's TLS certificate authenticates its host name.

As well as its normal function of providing an association between a domain name and a certificate, we are using the existance of a TLSA record to signal to the client that it can expect a valid server certificate.

The protocol described in this memo adds new security checks that can cause email delivery to be delayed when a security failure is detected.

2. Terminology

ADMD:

An ADministrative Management Domain, as described in the Internet Mail Architecture [RFC5598].

SMTP server host name:

The target of a (possibly implicit) MX record.

Inter-domain SMTP:

SMTP between different ADMDs across the public Internet, where a client sends mail to a publicly-referenced SMTP server.

Mail domain:

The part of an email address after the "@"; also the owner name of a (possibly implicit) MX record.

MX resolution:

The algorithm for resolving a mail domain into a set of SMTP server hosts, described in [RFC5321] section 5.

Publicly-referenced SMTP server:

An SMTP server which runs on port 25 of an Internet host located using MX resolution. (This term is from [RFC3207].)

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this memo are to be interpreted as described in [RFC2119].

3. Details of SMTP with TLSA

In the following we describe some additions to the usual MX resolution algorithm described in [RFC5321] section 5. If there is any conflict between [RFC5321] and this memo, that is an error in this memo.

The client SHALL look up the MX RRset for the mail domain. There are three succesful results that yield a list of SMTP server host names:

• A list of one or more MX records;
• An implicit MX record, in lieu of an empty list of MX records;
• A CNAME to a successful result.

If the lookup is not successful, the client SHALL proceed as usual.

All of these DNS RRsets MUST be "secure" according to DNSSEC validation ([RFC4033] section 5). In the case of an implicit MX record, there MUST be a secure denial of existence of an MX RRset for the mail domain. In the case of a (chain of) CNAME RRs, all the CNAMEs MUST be secure as well as their ultimate target.

If any of the responses is "bogus", the client MUST treat this as a temporary error.

If these security requirements are not satisfied, this protocol does not take effect. The client SHOULD fall back to insecure delivery (which might be over unauthenticated TLS).

The client now has an authentic list of SMTP server host names and priority values. It processes this list as usual.

The rest of this section applies to each SMTP server host name individually.

When connecting to a server, the client SHALL look up its TLSA RRset as described in [I-D.ietf-dane-protocol] section 3. That is, the TLSA RRset owner name SHALL be "_25._tcp.hostname" where "hostname" is the SMTP server host name. The response can be one of the following (as listed in [I-D.ietf-dane-protocol] section 4.1):

• A secure answer containing one or more TLSA records, in which case the client SHALL proceed as descrbed below.
• A bogus answer, which the client SHALL treat as a temporary error.
• In the other cases the client SHOULD deliver to this server insecurely (which might be over unauthenticated TLS).

The client now has one or more TLSA records for the server it is connecting to.

The client MUST ensure that the server offers the STARTTLS service extension [RFC3207] in its response to the client's EHLO command ([RFC5321] section 4.1.1.1).

The client SHALL then issue the STARTTLS command which MUST be successful. It then proceeds with TLS negotiation.

The client SHALL validate the server's certificate as described in [I-D.ietf-dane-protocol] section 2.1.

The client SHALL verify the server's identity as described in [RFC6125] section 6. Its list of reference identifiers MUST include the SMTP server host name with type DNS-ID, and MAY include a second copy of the host name with type CN-ID.

If any of these checks fail, the client MUST disconnect from the server and treat this as a temporary failure.

The client can now proceed to deliver mail securely.

The client MAY wish to insert a header at the start of the message to record the fact that it authenticated the server. XXX: Perhaps the form of this header should be specified here.

4. IANA Considerations

No IANA action is required.

5. Security considerations

This memo provides only conditional security. It allows a server to publish in the DNS the details of how it can be authenticated. Clients that implement this protocol can use it to provide a strong guarantee that they are sending mail to the correct place.

There is no way for a server to tell if a client has authenticated it using this protocol, since SMTP has no mechanism to signal this information.

We do not specify that clients check that all of a mail domain's SMTP server host names consistently have or do not have TLSA records. This is so that partial or incremental deployment does not break mail delivery. Inconsistencies are likely if a domain has a third-party backup MX, for example.

We do not specify that clients check the DNSSEC state of the SMTP server address records. This is not necessary since the certificate checks ensure that the client has connected to the correct server. (The address records will normally have the same security state as the TLSA records, but they can differ if there are CNAME or DNAME indirections.)

This memo does not make any changes to SMTP client authentication. Inter-domain SMTP client authentication remains extremely weak.

6. References

6.1. Normative References

6.2. Informative References

Appendix A. Rationale - choice of certificate identity

There are a number of reasons for the certificate to authenticate the SMTP server host name rather than the mail domain.

SMTP allows a client to transfer mail to recipients at multiple domains in the same connection. If the certificate identifies the host name then it does not need to list all the possible mail domains.

It is not in general feasible for the server to select a mail domain certificate based on the recipient domains when the connection is established (using Server Name Indication, [RFC6066] section 3), because an SMTP client might not know all of the recipients when it establishes the connection.

Outgoing SMTP relays and message submission servers handle mail for any domain, so in those cases the only sensible option is for the certificate to contain the host name. It is more consistent for incoming MX server certificates to match.

It is common for SMTP servers to act in multiple roles, as outgoing relays or as incoming MX servers, depending on the client identity. It is simpler if the server can present the same certificate regardless of the role in which it is to act.

Sometimes the server does not know its role until the client has authenticated, which usually occurs after TLS has been established.

Author's Address