RFC 5451
Message Header Field for Indicating Message Authentication Status
Network Working Group M. Kucherawy Request for Comments: 5451 Sendmail, Inc. Category: Standards Track April 2009 Message Header Field for Indicating Message Authentication Status Status of This Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited. Copyright Notice Copyright (c) 2009 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 in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.Abstract
This memo defines a new header field for use with electronic mail messages to indicate the results of message authentication efforts. Any receiver-side software, such as mail filters or Mail User Agents (MUAs), may use this message header field to relay that information in a convenient way to users or to make sorting and filtering decisions.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1. Purpose . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2. Trust Boundary . . . . . . . . . . . . . . . . . . . . . . 4 1.3. Processing Scope . . . . . . . . . . . . . . . . . . . . . 5 1.4. Requirements . . . . . . . . . . . . . . . . . . . . . . . 5 1.5. Definitions . . . . . . . . . . . . . . . . . . . . . . . 5 1.5.1. General . . . . . . . . . . . . . . . . . . . . . . . 5 1.5.2. Security . . . . . . . . . . . . . . . . . . . . . . . 5 1.5.3. Email Architecture . . . . . . . . . . . . . . . . . . 6 1.6. Trust Environment . . . . . . . . . . . . . . . . . . . . 7 2. Definition and Format of the Header Field . . . . . . . . . . 8 2.1. General Description . . . . . . . . . . . . . . . . . . . 8 2.2. Formal Definition . . . . . . . . . . . . . . . . . . . . 8 2.3. Authentication Identifier Field . . . . . . . . . . . . . 10 2.4. Result Values . . . . . . . . . . . . . . . . . . . . . . 12 2.4.1. DKIM and DomainKeys Results . . . . . . . . . . . . . 12 2.4.2. SPF and Sender-ID Results . . . . . . . . . . . . . . 13 2.4.3. "iprev" Results . . . . . . . . . . . . . . . . . . . 14 2.4.4. SMTP AUTH Results . . . . . . . . . . . . . . . . . . 14 2.4.5. Extension Result Codes . . . . . . . . . . . . . . . . 15 2.5. Authentication Methods . . . . . . . . . . . . . . . . . . 15 2.5.1. Definition of Initial Methods . . . . . . . . . . . . 16 2.5.2. Extension Methods . . . . . . . . . . . . . . . . . . 16 3. The "iprev" Authentication Method . . . . . . . . . . . . . . 17 4. Adding the Header Field to A Message . . . . . . . . . . . . . 18 4.1. Header Field Position and Interpretation . . . . . . . . . 19 4.2. Local Policy Enforcement . . . . . . . . . . . . . . . . . 20 5. Removing the Header Field . . . . . . . . . . . . . . . . . . 20 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22 6.1. The Authentication-Results Header Field . . . . . . . . . 22 6.2. Email Authentication Method Name Registry . . . . . . . . 22 6.3. Email Authentication Result Name Registry . . . . . . . . 24 7. Security Considerations . . . . . . . . . . . . . . . . . . . 26 7.1. Forged Header Fields . . . . . . . . . . . . . . . . . . . 26 7.2. Misleading Results . . . . . . . . . . . . . . . . . . . . 27 7.3. Header Field Position . . . . . . . . . . . . . . . . . . 28 7.4. Reverse IP Query Denial-of-Service Attacks . . . . . . . . 28 7.5. Mitigation of Backscatter . . . . . . . . . . . . . . . . 28 7.6. Internal MTA Lists . . . . . . . . . . . . . . . . . . . . 28 7.7. Attacks against Authentication Methods . . . . . . . . . . 28 7.8. Intentionally Malformed Header Fields . . . . . . . . . . 29 7.9. Compromised Internal Hosts . . . . . . . . . . . . . . . . 29 7.10. Encapsulated Instances . . . . . . . . . . . . . . . . . . 29 7.11. Reverse Mapping . . . . . . . . . . . . . . . . . . . . . 29
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 30 8.1. Normative References . . . . . . . . . . . . . . . . . . . 30 8.2. Informative References . . . . . . . . . . . . . . . . . . 30 Appendix A. Legacy MUAs . . . . . . . . . . . . . . . . . . . . . 32 Appendix B. Authentication-Results Examples . . . . . . . . . . . 33 B.1. Trivial Case; Header Field Not Present . . . . . . . . . . 33 B.2. Nearly Trivial Case; Service Provided, But No Authentication Done . . . . . . . . . . . . . . . . . . . 34 B.3. Service Provided, Authentication Done . . . . . . . . . . 35 B.4. Service Provided, Several Authentications Done, Single MTA . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 B.5. Service Provided, Several Authentications Done, Different MTAs . . . . . . . . . . . . . . . . . . . . . . 37 B.6. Service Provided, Multi-Tiered Authentication Done . . . . 39 Appendix C. Operational Considerations about Message Authentication . . . . . . . . . . . . . . . . . . . 41 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . 431. Introduction
This memo defines a new header field for electronic mail messages that presents the results of a message authentication effort in a machine-readable format. The intent is to create a place to collect such data when message authentication mechanisms are in use so that a Mail User Agent (MUA) and downstream filters can make filtering decisions and/or provide a recommendation to the user as to the validity of the message's origin and possibly the integrity of its content. End users are not expected to be direct consumers of this header field. This header field is intended for consumption by programs that will then use or render such data in a human-usable form. This memo defines both the format of this new header field and discusses the implications of its presence or absence. However, it does not discuss how the data contained in the header field should be used (i.e. what filtering decisions are appropriate, or how an MUA might render these results) as these are local policy and/or user interface design questions that are not appropriate for this memo. At the time of publication of this memo, [AUTH], [DKIM], [DOMAINKEYS], [SENDERID], and [SPF] are published DNS domain-level email authentication methods in common use. This proposal is not intended to be restricted to domain-based authentication, but this has proven to be a good starting point for implementations. As various methods emerge, it is necessary to prepare for their appearance and encourage convergence in the area of interfacing verifiers to filters and MUAs.
Although [SPF] defined a header field called Received-SPF and [DOMAINKEYS] defined one called DomainKey-Status for this purpose, those header fields are specific to the conveyance of their respective results only and thus are insufficient to satisfy the requirements enumerated below.1.1. Purpose
The header field defined in this memo is expected to serve several purposes: 1. Convey the results of various message authentication checks being applied by upstream filters and Mail Transfer Agents (MTAs) to MUAs and downstream filters within the same "trust domain", as such agents may wish to render those results to end users or use that data to apply more or less stringent content checks based on authentication results; 2. Provide a common location within a message for this data; 3. Create an extensible framework for reporting new authentication methods as they emerge. In particular, the mere presence of this header field should not be construed as meaning that its data is valid, but rather that it is asserting validity based on one or more authentication schemes somewhere upstream. For an MUA or downstream filter to treat the assertions as actually valid, there must be an assessment of the trust relationship between such agents and the validating MTA.1.2. Trust Boundary
This document makes several references to the "trust boundary" of an administrative management domain (ADMD). Given the diversity among existing mail environments, a precise definition of this term isn't possible. Simply put, a transfer from the creator of the header field to the consumer must occur within a context of trust that the creator's information is correct. How this trust is obtained is outside the scope of this document. It is entirely a local matter. Thus, this document defines a "trust boundary" as the delineation between "external" and "internal" entities; "external" here includes all hosts that do not deliberately provide some kind of messaging service for the receiving ADMD's users, and "internal" includes those hosts that do. By this definition, the hosts within a "trust boundary" may lie entirely within a receiving ADMD's direct control,
or they can include hosts managed by another ADMD (such as an ISP or commercial filtering service) but that also provide services for the former.1.3. Processing Scope
This proposal is intended to address the needs of authenticating messages or properties of messages during their actual transport. It is not meant to address the security of messages that might be encapsulated within other messages, such as a message/rfc822 [MIME] part within a message.1.4. Requirements
This memo establishes no new requirements on existing protocols or servers. In particular, this memo establishes no requirement on MTAs to reject or filter arriving messages that do not pass authentication checks. The data conveyed by the defined header field's contents are for the information of MUAs and filters and should be used at their discretion.1.5. Definitions
This section defines various terms used throughout this document.1.5.1. General
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 [KEYWORDS].1.5.2. Security
[SECURITY] discusses authentication and authorization and the conflation of the two concepts. The use of those terms within the context of recent message-security work has given rise to slightly different definitions, and this document reflects those current usages, as follows: o "Authorization" is the establishment of permission to use a resource or represent an identity. In this context, authorization indicates that a message from a particular ADMD arrived via a route the ADMD has explicitly approved.
o "Authentication" is the assertion of validity of a piece of data
about a message (such as the sender's identity) or the message in
its entirety.
As examples: [SPF] and [SENDERID] are authorization mechanisms in
that they express a result that shows whether or not the ADMD that
apparently sent the message has explicitly authorized the connecting
[SMTP] client to relay messages on its behalf but do not actually
validate any property of the message itself. By contrast, [DKIM] is
agnostic as to the routing of a message but uses cryptographic
signatures to authenticate agents claiming responsibility for the
message (which implies authorization) and ensure it was not modified
in transit. Since the signatures are not tied to SMTP connections,
they can be added by either the ADMD of origin, intermediate ADMDs
(such as a mailing list server), or both.
Rather than create a separate header field for each class of
solution, this proposal groups them both into a single header field.
1.5.3. Email Architecture
o A "border MTA" is an MTA that acts as a gateway between the
general Internet and the users within an organizational boundary.
(See also Section 1.2.)
o A "delivery MTA" (or Mail Delivery Agent or MDA) is an MTA that
actually enacts delivery of a message to a user's inbox or other
final delivery.
o An "intermediate MTA" is an MTA that handles messages after a
border MTA and before a delivery MTA.
The following diagram illustrates the flow of mail among these
defined components:
+-----+ +-----+ +------------+
| MUA |-->| MSA |-->| Border MTA |
+-----+ +-----+ +------------+
|
|
V
+----------+
| Internet |
+----------+
|
|
V
+-----+ +-----+ +------------------+ +------------+
| MUA |<--| MDA |<--| Intermediate MTA |<--| Border MTA |
+-----+ +-----+ +------------------+ +------------+
Generally, it is assumed that the work of applying message
authentication schemes takes place at a border MTA or a delivery MTA.
This specification is written with that assumption in mind. However,
there are some sites at which the entire mail infrastructure consists
of a single host. In such cases, such terms as "border MTA" and
"delivery MTA" may well apply to the same machine or even the very
same agent. It is also possible that some message authentication
tests could take place on an intermediate MTA. Although this
document doesn't specifically describe such cases, they are not meant
to be excluded from this specification.
See [EMAIL-ARCH] for further discussion on general email system
architecture, and Appendix C of this memo for discussion about the
common aspects of email authentication in current environments.
1.6. Trust Environment
This new header field permits one or more message validation
mechanisms to communicate its output to one or more separate
assessment mechanisms. These mechanisms operate within a unified
trust boundary that defines an Administrative Management Domain
(ADMD). An ADMD contains one or more entities that perform
validation and generate the header field, and one or more that
consume it for some type of assessment. The field contains no
integrity or validation mechanism of its own, so its presence must be
trusted implicitly. Hence, use of the header field depends upon
ensuring that mail entering the ADMD has instances of the header
field claiming to be valid within its boundaries removed, so that
occurrences of such header fields can be used safely by consumers.
The "authserv-id" token defined in Section 2.2 can be used to label an entire ADMD or a specific validation engine within an ADMD. Although the labeling scheme is left as an operational choice, some guidance for selecting a token is provided within this proposal.2. Definition and Format of the Header Field
This section gives a general overview of the format of the header field being defined, and then provides more formal specification.2.1. General Description
The new header field being defined here is called "Authentication- Results". It is a Structured Header Field as defined in [MAIL] and thus all of the related definitions in that document apply. This new header field SHOULD be added at the top of the message as it transits MTAs that do authentication checks so some idea of how far away the checks were done can be inferred. It therefore should be treated as a Trace Field as defined in [MAIL], and thus all of the related definitions in that document apply. The value of the header field (after removing [MAIL] comments) consists of an authentication identifier, an optional version, and then a series of "method=result" statements indicating which authentication method(s) were applied and their respective results, and then, for each applied method, an optional "reason" string plus optional "property=value" statements indicating which message properties were evaluated to reach that conclusion. The header field MAY appear more than once in a single message, or more than one result MAY be represented in a single header field, or a combination of these MAY be applied.2.2. Formal Definition
Formally, the header field is specified as follows using [ABNF]: authres-header = "Authentication-Results:" [CFWS] authserv-id [ CFWS version ] ( [CFWS] ";" [CFWS] "none" / 1*resinfo ) [CFWS] CRLF ; the special case of "none" is used to indicate that no ; message authentication is performed authserv-id = dot-atom ; see below for a description of this element
version = 1*DIGIT [CFWS]
; indicates which version of this specification is in use;
; this specification is version "1"; the absence of a
; version implies this version of the specification
resinfo = [CFWS] ";" methodspec [ CFWS reasonspec ]
*( CFWS propspec )
methodspec = [CFWS] method [CFWS] "=" [CFWS] result
; indicates which authentication method was evaluated
reasonspec = "reason" [CFWS] "=" [CFWS] value
; a free-form comment on the reason the given result
; was returned
propspec = ptype [CFWS] "." [CFWS] property [CFWS] "=" pvalue
; an indication of which properties of the message
; were evaluated by the authentication scheme being
; applied to yield the reported result and would be
; useful to reveal to end users as authenticated
method = dot-atom [ [CFWS] "/" [CFWS] version ]
; a method indicates which method's result is
; represented by "result", and is one of the methods
; explicitly defined as valid in this document
; or is an extension method as defined below
result = dot-atom
; indicates the results of the attempt to authenticate
; the message; see below for details
ptype = "smtp" / "header" / "body" / "policy"
; indicates whether the property being evaluated was
; a parameter to an [SMTP] command, or was a value taken
; from a message header field, or was some property of
; the message body, or some other property evaluated by
; the receiving MTA
property = dot-atom
; if "ptype" is "smtp", this indicates which [SMTP]
; command provided the value that was evaluated by the
; authentication scheme being applied; if "ptype" is
; "header", this indicates from which header field the
; value being evaluated was extracted; if "ptype" is
; "body", this indicates the offset into the body at which
; content of interest was detected; if "ptype" is "policy"
; then this indicates the name of the policy that caused
; this header field to be added (see below)
pvalue = [CFWS] ( value / [ [ local-part ] "@" ] domain-name )
[CFWS]
; the value extracted from the message property defined
; by the "ptype.property" construction; if the value
; identifies something intended to be an e-mail identity,
; then it MUST use the right hand portion of this ABNF
; definition
The "local-part" is as defined in Section 3.4.1, and "dot-atom" is
defined in Section 3.2.3, of [MAIL].
The "value" is as defined in Section 5.1 of [MIME].
The "domain-name" is as defined in Section 3.5 of [DKIM].
The "dot-atom" used in a "result" above is further constrained by the
necessity of being enumerated in Section 2.4 or an amendment to it.
See Section 2.3 for a description of the "authserv-id" element.
The list of commands eligible for use with the "smtp" ptype can be
found in [SMTP] and subsequent amendments.
"CFWS" is as defined in Section 3.2.2 of [MAIL].
The "propspec" may be omitted if, for example, the method was unable
to extract any properties to do its evaluation yet has a result to
report.
The "ptype" and "property" values used by each authentication method
should be defined in the specification for that method (or its
amendments).
The "ptype" and "property" are case-insensitive.
A "ptype" value of "policy" indicates a policy decision about the
message not specific to a property of the message that could be
extracted. For example, if a method would normally report a
"ptype.property" of "header.From" and no From: header field was
present, the method can use "policy" to indicate that no conclusion
about the authenticity of the message could be reached.
2.3. Authentication Identifier Field
Every Authentication-Results header field has an authentication
identifier field ("authserv-id" above). This is similar in syntax to
a fully-qualified domain name.
The authentication identifier field provides a unique identifier that refers to the authenticating service within a given ADMD. The uniqueness of the identifier MUST be guaranteed by the ADMD that generates it and must pertain to exactly that one ADMD. This identifier is intended to be machine-readable and not necessarily meaningful to users. MUAs or downstream filters SHOULD use this identifier to determine whether or not the data contained in an Authentication-Results header field should be used. For simplicity and scalability, the authentication identifier SHOULD be a common token used throughout the ADMD, such as the DNS domain name used by or within that ADMD. For tracing and debugging purposes, the authentication identifier MAY instead be the hostname of the MTA performing the authentication check whose result is being reported. This is also useful for another purpose, as described in Section 4. Moreover, some implementations have considered appending a delimiter such as "/" and following it with useful transport tracing data such as the [SMTP] queue ID or a timestamp. It should be noted, however, that using a local, relative identifier like a single hostname, rather than a hierarchical and globally unique ADMD identifier like a DNS domain name, makes configuration more difficult for large sites. The hierarchical identifier permits aggregating related, trusted systems together under a single, parent identifier, which in turn permits assessing the trust relationship with a single reference. The alternative is a flat namespace requiring individually listing each trusted system. Since consumers must use the identifier to determine whether to use the contents of the header field: o Changes to the identifier impose a large, centralized administrative burden. o Ongoing administrative changes require constantly updating this centralized table, making it difficult to ensure that an MUA or downstream filter will have access to accurate information for assessing the usability of the header field's content. In particular, consumers of the header field will need to know not only the current identifier(s) in use, but previous ones as well to account for delivery latency or later re-assessment of the header field's contents. Examples of valid authentication identifiers are "example.com", "mail.example.org", "ms1.newyork.example.com", and "example-auth".
2.4. Result Values
Each individual authentication method returns one of a set of specific result values. The subsections below define these results for the authentication methods specifically supported by this memo, and verifiers SHOULD use these values as described below. New methods not specified in this document intended to be supported by the header field defined in this memo MUST include a similar result table either in its defining memo or in a supplementary one.2.4.1. DKIM and DomainKeys Results
The result values used by [DKIM] and [DOMAINKEYS] are as follows: none: The message was not signed. pass: The message was signed, the signature or signatures were acceptable to the verifier, and the signature(s) passed verification tests. fail: The message was signed and the signature or signatures were acceptable to the verifier, but they failed the verification test(s). policy: The message was signed but the signature or signatures were not acceptable to the verifier. neutral: The message was signed but the signature or signatures contained syntax errors or were not otherwise able to be processed. This result SHOULD also be used for other failures not covered elsewhere in this list. temperror: The message could not be verified due to some error that is likely transient in nature, such as a temporary inability to retrieve a public key. A later attempt may produce a final result. permerror: The message could not be verified due to some error that is unrecoverable, such as a required header field being absent. A later attempt is unlikely to produce a final result. A signature is "acceptable to the verifier" if it passes local policy checks (or there are no specific local policy checks). For example, a verifier might require that the signature(s) on the message be added using the DNS domain present in the From: header field of the message, thus making third-party signatures unacceptable.
[DKIM] advises that if a message fails verification, it should be treated as an unsigned message. A report of "fail" here permits the receiver of the report to decide how to handle the failure. A report of "neutral" or "none" preempts that choice, ensuring the message will be treated as if it had not been signed.2.4.2. SPF and Sender-ID Results
The result values are used by [SPF] and [SENDERID] as follows: none: No policy records were published at the sender's DNS domain. neutral: The sender's ADMD has asserted that it cannot or does not want to assert whether or not the sending IP address is authorized to send mail using the sender's DNS domain. pass: The client is authorized by the sender's ADMD to inject or relay mail on behalf of the sender's DNS domain. policy: The client is authorized to inject or relay mail on behalf of the sender's DNS domain according to the authentication method's algorithm, but local policy dictates that the result is unacceptable. hardfail: This client is explicitly not authorized to inject or relay mail using the sender's DNS domain. softfail: The sender's ADMD believes the client was not authorized to inject or relay mail using the sender's DNS domain, but is unwilling to make a strong assertion to that effect. temperror: The message could not be verified due to some error that is likely transient in nature, such as a temporary inability to retrieve a policy record from DNS. A later attempt may produce a final result. permerror: The message could not be verified due to some error that is unrecoverable, such as a required header field being absent or a syntax error in a retrieved DNS TXT record. A later attempt is unlikely to produce a final result. The distinction between and interpretation of "none" and "neutral" under these methods is discussed further in [SPF]. The "policy" result would be returned if, for example, [SPF] returned as "pass" result, but a local policy check matches the sending DNS domain to one found in an explicit list of unacceptable DNS domains (e.g., spammers).
If the retrieved sender policies used to evaluate [SPF] and [SENDERID] do not contain explicit provisions for authenticating the local-part (see Section 3.4.1 of [MAIL]) of an address, the "pvalue" reported along with results for these mechanisms SHOULD NOT include the local-part.2.4.3. "iprev" Results
The result values are used by the "iprev" method, defined in Section 3, are as follows: pass: The DNS evaluation succeeded, i.e., the "reverse" and "forward" lookup results were returned and were in agreement. fail: The DNS evaluation failed. In particular, the "reverse" and "forward" lookups each produced results but they were not in agreement, or the "forward" query completed but produced no result, e.g., a DNS RCODE of 3, commonly known as NXDOMAIN, or an RCODE of 0 (NOERROR) in a reply containing no answers, was returned. temperror: The DNS evaluation could not be completed due to some error that is likely transient in nature, such as a temporary DNS error, e.g., a DNS RCODE of 2, commonly known as SERVFAIL, or other error condition resulted. A later attempt may produce a final result. permerror: The DNS evaluation could not be completed because no PTR data are published for the connecting IP address, e.g., a DNS RCODE of 3, commonly known as NXDOMAIN, or an RCODE of 0 (NOERROR) in a reply containing no answers, was returned. This prevented completion of the evaluation. There is no "none" for this method since any TCP connection delivering email has an IP address associated with it, so some kind of evaluation will always be possible. For discussion of the format of DNS replies, see [DNS].2.4.4. SMTP AUTH Results
The result values are used by the [AUTH] method are as follows: none: SMTP authentication was not attempted. pass: The SMTP client had authenticated to the server reporting the result using the protocol described in [AUTH].
fail: The SMTP client had attempted to authenticate to the server
using the protocol described in [AUTH] but was not successful, yet
continued to send the message about which a result is being
reported.
temperror: The SMTP client attempted to authenticate using the
protocol described in [AUTH] but was not able to complete the
attempt due to some error which is likely transient in nature,
such as a temporary Lightweight Directory Access Protocol (LDAP)
lookup error. A later attempt may produce a final result.
permerror: The SMTP client attempted to authenticate using the
protocol described in [AUTH] but was not able to complete the
attempt due to some error that is likely not transient in nature,
such as a permanent LDAP lookup error. A later attempt is not
likely produce a final result.
Note that an agent making use of the data provided by this header
field SHOULD consider "fail" and "temperror" to be the synonymous in
terms of message authentication, i.e., the client did not
authenticate.
2.4.5. Extension Result Codes
Additional result codes (extension results) might be defined in the
future by later revisions or extensions to this specification.
Extension results beginning with "x-" will never be defined as
standard fields; such names are reserved for experimental use.
Result codes not beginning with "x-" MUST be registered with the
Internet Assigned Numbers Authority (IANA) and published in an RFC.
See Section 6 for further details.
Implementations reporting new result codes MUST use the "x-" prefix
until such time as the new method is registered by IANA.
Extension results MUST only be used within ADMDs that have explicitly
consented to use them. These results and the parameters associated
with them are not documented in RFCs. Therefore, they are subject to
change at any time and not suitable for production use. Any MTA, MUA
or downstream filter intended for production use SHOULD ignore or
delete any Authentication-Results header field that includes an
extension result.
2.5. Authentication Methods
This section defines the supported authentication methods and
discusses the proper means for applying experimental and other
extension methods.
2.5.1. Definition of Initial Methods
As they are currently existing specifications for message authentication, it is appropriate to define an authentication method identifier for each of [AUTH], [DKIM], [DOMAINKEYS], [SENDERID], and [SPF]. Therefore, the authentication method identifiers "auth", "dkim", "domainkeys", "sender-id", and "spf", respectively are hereby defined for MTAs applying those specifications for email message authentication. Furthermore, method "iprev" is defined in Section 3. See Section 6 for details.2.5.2. Extension Methods
Additional authentication method identifiers (extension methods) may be defined in the future by later revisions or extensions to this specification. Extension methods beginning with "x-" will never be defined as standard fields; such names are reserved for experimental use. Method identifiers not beginning with "x-" MUST be registered with the Internet Assigned Numbers Authority (IANA) and published in an RFC. See Section 6 for further details. Extension methods may be defined for the following reasons: 1. To allow additional information from new authentication systems to be communicated to MUAs or downstream filters. The names of such identifiers should reflect the name of the method being defined, but should not be needlessly long. 2. To allow the creation of "sub-identifiers" that indicate different levels of authentication and differentiate between their relative strengths, e.g., "auth1-weak" and "auth1-strong". Implementations of new methods MUST use the "x-" prefix until such time as the new method is registered by IANA. Authentication method implementors are encouraged to provide adequate information, via [MAIL] comments if necessary, to allow an MUA developer to understand or relay ancillary details of authentication results. For example, if it might be of interest to relay what data was used to perform an evaluation, such information could be relayed as a comment in the header field, such as: Authentication-Results: example.com; foo=pass bar.baz=blob (2 of 3 tests OK)
Experimental method identifiers MUST only be used within ADMDs that have explicitly consented to use them. These method identifiers and the parameters associated with them are not documented in RFCs. Therefore, they are subject to change at any time and not suitable for production use. Any MTA, MUA, or downstream filter intended for production use SHOULD ignore or delete any Authentication-Results header field that includes an experimental method identifier.3. The "iprev" Authentication Method
This section defines an additional authentication method called "iprev". In general, "iprev" is an attempt to verify that a client appears to be valid based on some DNS queries. Upon receiving a session initiation of some kind from a client, the IP address of the client peer is queried for matching names (i.e., a number-to-name translation, also known as a "reverse lookup" or a "PTR" record query). Once that result is acquired, a lookup of each of the names (i.e., a name-to-number translation, or an "A" or "AAAA" record query) thus retrieved is done. The response to this second check should result in at least one mapping back to the client's IP address. More algorithmically: if the client peer's IP address is I, the list of names to which I maps (after a "PTR" query) is the set N, and the union of IP addresses to which each member of N maps (after corresponding "A" and "AAAA" queries) is L, then this test is successful if I is an element of L. The response to a PTR query could contain multiple names. To prevent heavy DNS loads, agents performing these queries MUST be implemented such that the number of names evaluated by generation of corresponding A or AAAA queries is finite, though it MAY be configurable by an administrator. As an example, Section 5.5 of [SPF] chose a limit of 10 for its implementation of this algorithm. [DNS-IP6] discusses the query formats for the IPv6 case. A successful test using this algorithm constitutes a result of "pass" since the ADMD in which the client's PTR claims it belongs has confirmed that claim by including corresponding data in its DNS domain. A failure to match constitutes a "fail". There is no case in which a "neutral" result can be returned. The remaining "temperror" and "permerror" cases refer, respectively, to temporary and permanent DNS query errors.
There is some contention regarding the wisdom and reliability of this test. For example, in some regions it can be difficult for this test ever to pass because the practice of arranging to match the forward and reverse DNS is infrequently observed. Therefore, the actual implementation details of how a verifier performs an "iprev" test are not specified here. The verifier MAY report a successful or failed "iprev" test at its discretion having done some kind of check of the validity of the connection's identity using DNS. It is incumbent upon an agent making use of the reported "iprev" result to understand what exactly that particular verifier is attempting to report. Extensive discussion of reverse DNS mapping and its implications can be found in [DNSOP-REVERSE]. In particular, it recommends that applications avoid using this test as a means of authentication or security. Its presence in this memo is not an endorsement, but is merely acknowledgement that the method remains common and provides the means to relay the results of that test.4. Adding the Header Field to A Message
This specification makes no attempt to evaluate the relative strengths of various message authentication methods that may become available. As such, the order of the presented authentication methods and results MUST NOT be used either to imply or infer the importance or strength of any given method over another. Instead, the MUA or downstream filter consuming this header field must interpret the result of each method based on its own knowledge of what that method evaluates. Each "method" MUST refer to an authentication method declared in the IANA registry, or an extension method as defined in Section 2.5.2, and each "result" MUST refer to a result code declared in the IANA registry, or an extension result code as defined in Section 2.4.5. See Section 6 for further information about the registered methods and result codes. An MTA compliant with this specification MUST add this header field (after performing one or more message authentication tests) to indicate which MTA or ADMD performed the test, which test got applied and what the result was. If an MTA applies more than one such test, it MUST add this header field either once per test, or once indicating all of the results. An MTA MUST NOT add a result to an existing header field. An MTA MAY add this header field containing only the authentication identifier portion to indicate explicitly that no message authentication schemes were applied prior to delivery of this message.
An MTA adding this header field must take steps to identify it as legitimate to the MUAs or downstream filters that will ultimately consume its content. One required process to do so is described in Section 5. Further measures may be required in some environments. Some possible solutions are enumerated in Section 7.1. This memo does not mandate any specific solution to this issue as each environment has its own facilities and limitations. For MTAs that add this header field, adding header fields in order (at the top), per Section 3.6 of [MAIL], is particularly important. Moreover, this header field SHOULD be inserted above any other trace header fields such MTAs might prepend. This allows easy detection of header fields that can be trusted. End users making direct use of this header field may inadvertently trust information that has not been properly vetted. If, for example, a basic [SPF] result were to be relayed that claims an authenticated addr-spec, the local-part of that addr-spec has actually not been authenticated. Thus, an MTA adding this header field SHOULD NOT include any data that has not been authenticated by the method(s) being applied. Moreover, MUAs SHOULD NOT render to users such information if it is presented by a method known not to authenticate it.4.1. Header Field Position and Interpretation
In order to ensure non-ambiguous results and avoid the impact of false header fields, MUAs and downstream filters SHOULD NOT interpret this header field unless specifically instructed to do so by the user or administrator. That is, this interpretation should not be "on by default". Naturally then, users or administrators should not activate such a feature unless they are certain the header field will be added by the border MTA that accepts the mail that is ultimately read by the MUA, and instances of the header field appearing to be from within the ADMD but actually added by foreign MTAs will be removed before delivery. Furthermore, MUAs and downstream filters SHOULD NOT interpret this header field unless the authentication identifier it bears appears to be one used within its own ADMD as configured by the user or administrator. MUAs and downstream filters MUST ignore any result reported using a "result" not specified in the result code registry, or a "ptype" not listed in the corresponding registry for such values as defined in Section 6. Moreover, such agents MUST ignore a result indicated for any "method" they do not specifically support.
An MUA SHOULD NOT reveal these results to end users unless the results are accompanied by, at a minimum, some associated reputation data about the authenticated origin identifiers within the message. For example, an attacker could register examp1e.com (note the digit "one") and send signed mail to intended victims; a verifier would detect that the signature was valid and report a "pass" even though it's clear the DNS domain name was intended to mislead. See Section 7.2 for further discussion. As stated in Section 2.1, this header field SHOULD be treated as though it were a trace header field as defined in Section 3.6.7 of [MAIL], and hence MUST NOT be reordered and MUST be prepended to the message, so that there is generally some indication upon delivery of where in the chain of handling MTAs the message authentication was done. MUAs SHOULD ignore instances of this header field discovered within message/rfc822 [MIME] attachments. Further discussion of this can be found in Section 7 below.4.2. Local Policy Enforcement
If a site's local policy is to consider a non-recoverable failure result (e.g., "fail" for DKIM, "hardfail" for SPF) for any particular authentication method as justification to reject the message completely, the border MTA SHOULD issue an [SMTP] rejection response to the message rather than adding this header field with the failure result and allowing it to proceed toward delivery. This is more desirable than allowing the message to reach an internal host's MTA or spam filter, thus possibly generating a local rejection such as a [DSN] to a forged originator. The same MAY also be done for local policy decisions overriding the results of the authentication methods (e.g., the "policy" result codes described in Section 2.4). Such rejections at the SMTP protocol level are not possible if local policy is enforced at the MUA and not the MTA. Unfortunately, this may be a common scenario.5. Removing the Header Field
For security reasons, any MTA conforming to this specification MUST delete any discovered instance of this header field that claims to have been added within its trust boundary and that did not come from another trusted MTA. For example, an MTA (border or otherwise) for example.com receiving a message MUST delete any instance of this
header field bearing an authentication identifier indicating the header field was added within example.com prior to adding its own header fields. This may mean each MTA will have to be equipped with a list of internal MTAs known to be compliant (and hence trustworthy). For simplicity and maximum security, a border MTA MAY remove all instances of this header field on mail crossing into its trust boundary. However, this may conflict with the desire to access authentication results performed by trusted external service providers. It may also invalidate signed messages whose signatures cover external instances of this header field. A more robust border MTA could allow a specific list of authenticating MTAs whose information should be let in, removing all others. As stated in Section 1.2, a formal definition of "trust boundary" is deliberately not made here. It is entirely possible that a border MTA for example.com might explicitly trust authentication results asserted by upstream host example.net even though they exist in completely disjoint administrative boundaries. In that case, the border MTA MAY elect not to delete those results; moreover, the upstream host doing some authentication work could apply a signing technology such as [DKIM] on its own results to assure downstream hosts of their authenticity. An example of this is provided in Appendix B. Similarly, in the case of messages signed using [DKIM] or other message signing methods that sign header fields, this may invalidate one or more signatures on the message if they covered the header field to be removed at the time of signing. This behavior can be desirable since there's little value in validating the signature on a message with forged headers. However, signing agents MAY therefore elect to omit these header fields from signing to avoid this situation. An MTA SHOULD remove any instance of this header field bearing a version (express or implied) that it does not support. However, an MTA MUST remove such a header if the [SMTP] connection relaying the message is not from a trusted internal MTA.
6. IANA Considerations
IANA has registered a new header field and created two new tables as described below.6.1. The Authentication-Results Header Field
Per [IANA-HEADERS], the "Authentication-Results" header field has been added to the IANA Permanent Message Header Field Registry. The following is the registration template: Header field name: Authentication-Results Applicable protocol: mail ([MAIL]) Status: Standard Author/Change controller: IETF Specification document(s): RFC 5451 Related information: Requesting review of any proposed changes and additions to this field is recommended.6.2. Email Authentication Method Name Registry
Names of message authentication methods supported by this specification must be registered with IANA, with the exception of experimental names as described in Section 2.5.2. New entries are assigned only for values that have been documented in a published RFC that has had IETF Review, per [IANA-CONSIDERATIONS]. Each method must register a name, the specification that defines it, one or more "ptype" values appropriate for use with that method, which "property" value(s) should be reported by that method, and a description of the "value" to be used with each.
The initial set of entries in this registry is as follows: +------------+----------+--------+----------------+--------------------+ | Method | Defined | ptype | property | value | +------------+----------+--------+----------------+--------------------+ | auth | RFC4954 | smtp | auth | AUTH parameter of | | | | | | the SMTP MAIL | | | | | | command | +------------+----------+--------+----------------+--------------------+ | dkim | RFC4871 | header | d | value of | | | | | | signature "d" tag | | | | +----------------+--------------------+ | | | | i | value of | | | | | | signature "i" tag | +------------+----------+--------+----------------+--------------------+ | domainkeys | RFC4870 | header | d | value of | | | | | | signature "d" tag | | | | +----------------+--------------------+ | | | | from | value of From | | | | | | header field after | | | | | | removing comments | | | | | | and local-part if | | | | | | not authenticated | | | | +----------------+--------------------+ | | | | sender | value of Sender | | | | | | header field after | | | | | | removing comments | | | | | | and local-part if | | | | | | not authenticated | +------------+----------+--------+----------------+--------------------+ | iprev | this | policy | iprev | client IP address | | | document | | | | +------------+----------+--------+----------------+--------------------+ | sender-id | RFC4406 | header | name of header | value of header | | | | | field used by | field used by PRA | | | | | the Purported | after removing | | | | | Responsible | comments and parts | | | | | Address (PRA) | not authenticated | +------------+----------+--------+----------------+--------------------+ | spf | RFC4408 | smtp | mailfrom | envelope sender | | | | | | after removing | | | | | | parts not | | | | | | authenticated | | | +--------+----------------+--------------------+ | | | smtp | helo | HELO/EHLO value | +------------+----------+--------+----------------+--------------------+
6.3. Email Authentication Result Name Registry
Names of message authentication result codes supported by this specification must be registered with IANA, with the exception of experimental codes as described in Section 2.4.5. New entries are assigned only for result codes that have been documented in a published RFC that has had IETF Review, per [IANA-CONSIDERATIONS]. Each code must register a name, the document that establishes the registration, the authentication method(s) that uses it, and either a definition of the semantics of its use or a reference to the place where those semantics are defined. The initial set of entries in this registry is as follows: +-----------+----------+----------------+------------------------------+ | Code | Defined | Auth Method(s) | Meaning | +-----------+----------+----------------+------------------------------+ | none | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | | | +----------------+------------------------------+ | | | auth | section 2.4.4 | +-----------+----------+----------------+------------------------------+ | pass | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | | | +----------------+------------------------------+ | | | iprev | section 2.4.3 | | | +----------------+------------------------------+ | | | auth | section 2.4.4 | +-----------+----------+----------------+------------------------------+ | fail | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | iprev | section 2.4.3 | | | +----------------+------------------------------+ | | | auth | section 2.4.4 | +-----------+----------+----------------+------------------------------+
| policy | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | +-----------+----------+----------------+------------------------------+ | neutral | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | +-----------+----------+----------------+------------------------------+ | temperror | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | | | +----------------+------------------------------+ | | | iprev | section 2.4.3 | | | +----------------+------------------------------+ | | | auth | section 2.4.4 | +-----------+----------+----------------+------------------------------+ | permerror | this | dkim | section 2.4.1 | | | document | domainkeys | | | | +----------------+------------------------------+ | | | spf | section 2.4.2 | | | | sender-id | | | | +----------------+------------------------------+ | | | iprev | section 2.4.3 | | | +----------------+------------------------------+ | | | auth | section 2.4.4 | +-----------+----------+----------------+------------------------------+ | hardfail | this | spf | section 2.4.2 | | | document | sender-id | | +-----------+----------+----------------+------------------------------+ | softfail | this | spf | section 2.4.2 | | | document | sender-id | | +-----------+----------+----------------+------------------------------+
7. Security Considerations
The following security considerations apply when adding or processing the "Authentication-Results" header field:7.1. Forged Header Fields
An MUA or filter that accesses a mailbox whose mail is handled by a non-conformant MTA, and understands Authentication-Results header fields, could potentially make false conclusions based on forged header fields. A malicious user or agent could forge a header field using the DNS domain of a receiving ADMD as the authserv-id token in the value of the header field, and with the rest of the value claim that the message was properly authenticated. The non-conformant MTA would fail to strip the forged header field, and the MUA could inappropriately trust it. It is for this reason an MUA should not have processing of the "Authentication-Results" header field enabled by default; instead it should be ignored, at least for the purposes of enacting filtering decisions, unless specifically enabled by the user or administrator after verifying that the border MTA is compliant. It is acceptable to have an MUA aware of this specification, but have an explicit list of hostnames whose "Authentication-Results" header fields are trustworthy; however, this list should initially be empty. Proposed alternate solutions to this problem are nascent: 1. Possibly the simplest is a digital signature protecting the header field, such as using [DKIM], that can be verified by an MUA by using a posted public key. Although one of the main purposes of this memo is to relieve the burden of doing message authentication work at the MUA, this only requires that the MUA learn a single authentication scheme even if a number of them are in use at the border MTA. Note that [DKIM] requires that the From header field be signed, although in this application, the signing agent (a trusted MTA) likely cannot authenticate that value, so the fact that it is signed should be ignored. 2. Another would be a means to interrogate the MTA that added the header field to see if it is actually providing any message authentication services and saw the message in question, but this isn't especially palatable given the work required to craft and implement such a scheme. 3. Yet another might be a method to interrogate the internal MTAs that apparently handled the message (based on Received: header
fields) to determine whether any of them conform to Section 5 of
this memo. This, too, has potentially high barriers-to-entry.
4. Extensions to [IMAP], [SMTP], and [POP3] could be defined to
allow an MUA or filtering agent to acquire the "authserv-id" in
use within an ADMD, thus allowing it to identify which
Authentication-Results header fields it can trust.
5. On the presumption that internal MTAs are fully compliant with
Section 3.6 of [MAIL], and the compliant internal MTAs are using
their own host names or the ADMD's DNS domain name as the
"authserv-id" token, the header field proposed here should always
appear above a Received: header added by a trusted MTA. This can
be used as a test for header field validity.
Support for some of these is planned for future work.
In any case, a mechanism needs to exist for an MUA or filter to
verify that the host that appears to have added the header field (a)
actually did so, and (b) is legitimately adding that header field for
this delivery. Given the variety of messaging environments deployed
today, consensus appears to be that specifying a particular mechanism
for doing so is not appropriate for this memo.
Mitigation of the forged header field attack can also be accomplished
by moving the authentication results data into meta-data associated
with the message. In particular, an [SMTP] extension could be
established which is used to communicate authentication results from
the border MTA to intermediate and delivery MTAs; the latter of these
could arrange to store the authentication results as meta-data
retrieved and rendered along with the message by an [IMAP] client
aware of a similar extension in that protocol. The delivery MTA
would be told to trust data via this extension only from MTAs it
trusts, and border MTAs would not accept data via this extension from
any source. There is no vector in such an arrangement for forgery of
authentication data by an outside agent.
7.2. Misleading Results
Until some form of service for querying the reputation of a sending
agent is widely deployed, the existence of this header field
indicating a "pass" does not render the message trustworthy. It is
possible for an arriving piece of spam or other undesirable mail to
pass checks by several of the methods enumerated above (e.g., a piece
of spam signed using [DKIM] by the originator of the spam, which
might be a spammer or a compromised system). In particular, this
issue is not resolved by forged header field removal discussed above.
Hence, MUAs and downstream filters must take some care with use of this header even after possibly malicious headers are scrubbed.7.3. Header Field Position
Despite the requirements of [MAIL], header fields can sometimes be reordered enroute by intermediate MTAs. The goal of requiring header field addition only at the top of a message is an acknowledgement that some MTAs do reorder header fields, but most do not. Thus, in the general case, there will be some indication of which MTAs (if any) handled the message after the addition of the header field defined here.7.4. Reverse IP Query Denial-of-Service Attacks
Section 5.5 of [SPF] describes a DNS-based denial-of-service attack for verifiers that attempt DNS-based identity verification of arriving client connections. A verifier wishing to do this check and report this information SHOULD take care not to go to unbounded lengths to resolve "A" and "PTR" queries. MUAs or other filters making use of an "iprev" result specified by this memo SHOULD be aware of the algorithm used by the verifier reporting the result and thus be aware of its limitations.7.5. Mitigation of Backscatter
Failing to follow the instructions of Section 4.2 can result in a denial-of-service attack caused by the generation of [DSN] messages (or equivalent) to addresses that did not send the messages being rejected.7.6. Internal MTA Lists
Section 5 describes a procedure for scrubbing headers that may contain forged authentication results about a message. A compliant installation will have to include, at each MTA, a list of other MTAs known to be compliant and trustworthy. Failing to keep this list current as internal infrastructure changes may expose an ADMD to attack.7.7. Attacks against Authentication Methods
If an attack becomes known against an authentication method, clearly then the agent verifying that method can be fooled into thinking an inauthentic message is authentic, and thus the value of this header field can be misleading. It follows that any attack against the authentication methods supported by this document (and later amendments to it) is also a security consideration here.
7.8. Intentionally Malformed Header Fields
It is possible for an attacker to add an Authentication-Results header field that is extraordinarily large or otherwise malformed in an attempt to discover or exploit weaknesses in header field parsing code. Implementors must thoroughly verify all such header fields received from MTAs and be robust against intentionally as well as unintentionally malformed header fields.7.9. Compromised Internal Hosts
An internal MUA or MTA that has been compromised could generate mail with a forged From header field and a forged Authentication-Results header field that endorses it. Although it is clearly a larger concern to have compromised internal machines than it is to prove the value of this header field, this risk can be mitigated by arranging that internal MTAs will remove this header field if it claims to have been added by a trusted border MTA (as described above), yet the [SMTP] connection is not coming from an internal machine known to be running an authorized MTA. However, in such a configuration, legitimate MTAs will have to add this header field when legitimate internal-only messages are generated. This is also covered in Section 5.7.10. Encapsulated Instances
[MIME] messages may contain attachments of type "message/rfc822", which contain other [MAIL] messages. Such an encapsulated message may also contain an Authentication-Results header field. Although the processing of these is outside of the intended scope of this document (see Section 1.3), some early guidance to MUA developers is appropriate here. Since MTAs are unlikely to strip Authentication-Results header fields after mailbox delivery, MUAs are advised in Section 4.1 to ignore such instances within [MIME] attachments. Moreover, when extracting a message digest to separate mail store messages or other media, such header fields should be removed so that they will never be interpreted improperly by MUAs that might later consume them.7.11. Reverse Mapping
Although Section 3 of this memo includes explicit support for the "iprev" method, its value as an authentication mechanism is limited. Implementors of both this proposal and agents that use the data it relays are encouraged to become familiar with the issues raised by [DNSOP-REVERSE] when deciding whether or not to include support for "iprev".
8. References
8.1. Normative References
[ABNF] Crocker, D. and P. Overell, "Augmented BNF for Syntax Specifications: ABNF", STD 68, RFC 5234, January 2008. [IANA-HEADERS] Klyne, G., Nottingham, M., and J. Mogul, "Registration Procedures for Message Header Fields", BCP 90, RFC 3864, September 2004. [KEYWORDS] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [MAIL] Resnick, P., Ed., "Internet Message Format", RFC 5322, October 2008. [MIME] Freed, N. and N. Borenstein, "Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies", RFC 2045, November 1996.8.2. Informative References
[AUTH] Siemborski, R. and A. Melnikov, "SMTP Service Extension for Authentication", RFC 4954, July 2007. [DKIM] Allman, E., Callas, J., Delany, M., Libbey, M., Fenton, J., and M. Thomas, "DomainKeys Identified Mail (DKIM) Signatures", RFC 4871, May 2007. [DNS] Mockapetris, P., "Domain names - implementation and specification", STD 13, RFC 1035, November 1987. [DNS-IP6] Thomson, S., Huitema, C., Ksinant, V., and M. Souissi, "DNS Extensions to Support IP Version 6", RFC 3596, October 2003. [DNSOP-REVERSE] Senie, D. and A. Sullivan, "Considerations for the use of DNS Reverse Mapping", Work in Progress, March 2008.
[DOMAINKEYS] Delany, M., "Domain-Based Email Authentication Using Public Keys Advertised in the DNS (DomainKeys)", RFC 4870, May 2007. [DSN] Moore, K. and G. Vaudreuil, "An Extensible Message Format for Delivery Status Notifications", RFC 3464, January 2003. [EMAIL-ARCH] Crocker, D., "Internet Mail Architecture", Work in Progress, October 2008. [IANA-CONSIDERATIONS] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. [IMAP] Crispin, M., "INTERNET MESSAGE ACCESS PROTOCOL - VERSION 4rev1", RFC 3501, March 2003. [POP3] Myers, J. and M. Rose, "Post Office Protocol - Version 3", STD 53, RFC 1939, May 1996. [SECURITY] Rescorla, E. and B. Korver, "Guidelines for Writing RFC Text on Security Considerations", BCP 72, RFC 3552, July 2003. [SENDERID] Lyon, J. and M. Wong, "Sender ID: Authenticating E-Mail", RFC 4406, April 2006. [SMTP] Klensin, J., "Simple Mail Transfer Protocol", RFC 5321, October 2008. [SPF] Wong, M. and W. Schlitt, "Sender Policy Framework (SPF) for Authorizing Use of Domains in E-Mail, Version 1", RFC 4408, April 2006.
Appendix A. Legacy MUAs
Implementors of this proposal should be aware that many MUAs are unlikely to be retrofitted to support the new header field and its semantics. In the interests of convenience and quicker adoption, a delivery MTA might want to consider adding things that are processed by existing MUAs in addition to the Authentication-Results header field. One suggestion is to include a Priority header field, on messages that don't already have such a header field, containing a value that reflects the strength of the authentication that was accomplished, e.g., "low" for weak or no authentication, "normal" or "high" for good or strong authentication. Some modern MUAs can already filter based on the content of this header field. However, there is keen interest in having MUAs make some kind of graphical representation of this header field's meaning to end users. Until this capability is added, other interim means of conveying authentication results may be necessary while this proposal and its successors are adopted.
Appendix B. Authentication-Results Examples
This section presents some examples of the use of this header field to indicate authentication results.B.1. Trivial Case; Header Field Not Present
The trivial case: Received: from mail-router.example.com (mail-router.example.com [192.0.2.1]) by server.example.org (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 From: sender@example.com Date: Fri, Feb 15 2002 16:54:30 -0800 To: receiver@example.org Message-Id: <12345.abc@example.com> Subject: here's a sample Hello! Goodbye! Example 1: Trivial case The "Authentication-Results" header field is completely absent. The MUA may make no conclusion about the validity of the message. This could be the case because the message authentication services were not available at the time of delivery, or no service is provided, or the MTA is not in compliance with this specification.
B.2. Nearly Trivial Case; Service Provided, But No Authentication Done
A message that was delivered by an MTA that conforms to this specification but provides no actual message authentication service: Authentication-Results: example.org; none Received: from mail-router.example.com (mail-router.example.com [192.0.2.1]) by server.example.org (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 From: sender@example.com Date: Fri, Feb 15 2002 16:54:30 -0800 To: receiver@example.org Message-Id: <12345.abc@example.com> Subject: here's a sample Hello! Goodbye! Example 2: Header present but no authentication done The "Authentication-Results" header field is present, showing that the delivering MTA conforms to this specification. It used its DNS domain name as the authserv-id. The presence of "none" (and the absence of any method and result tokens) indicates that no message authentication was done.
B.3. Service Provided, Authentication Done
A message that was delivered by an MTA that conforms to this specification and applied some message authentication: Authentication-Results: example.com; spf=pass smtp.mailfrom=example.net Received: from dialup-1-2-3-4.example.net (dialup-1-2-3-4.example.net [192.0.2.200]) by mail-router.example.com (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 From: sender@example.net Date: Fri, Feb 15 2002 16:54:30 -0800 To: receiver@example.com Message-Id: <12345.abc@example.net> Subject: here's a sample Hello! Goodbye! Example 3: Header reporting results The "Authentication-Results" header field is present, indicating that the border MTA conforms to this specification. The authserv-id is once again the DNS domain name. Furthermore, the message was authenticated by that MTA via the method specified in [SPF]. Note that since that method cannot authenticate the local-part, it has been omitted from the result's value. The MUA could extract and relay this extra information if desired.
B.4. Service Provided, Several Authentications Done, Single MTA
A message that was relayed inbound via a single MTA that conforms to this specification and applied three different message authentication checks: Authentication-Results: example.com; auth=pass (cram-md5) smtp.auth=sender@example.com; spf=pass smtp.mailfrom=example.com Authentication-Results: example.com; sender-id=pass header.from=example.com Received: from dialup-1-2-3-4.example.net (8.11.6/8.11.6) (dialup-1-2-3-4.example.net [192.0.2.200]) by mail-router.example.com (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 Date: Fri, Feb 15 2002 16:54:30 -0800 To: receiver@example.net From: sender@example.com Message-Id: <12345.abc@example.com> Subject: here's a sample Hello! Goodbye! Example 4: Headers reporting results from one MTA The "Authentication-Results" header field is present, indicating the delivering MTA conforms to this specification. Once again, the receiving DNS domain name is used as the authserv-id. Furthermore, the sender authenticated herself/himself to the MTA via a method specified in [AUTH], and both [SPF] and [SENDERID] checks were done and passed. The MUA could extract and relay this extra information if desired. Two "Authentication-Results" header fields are not required since the same host did all of the checking. The authenticating agent could have consolidated all the results into one header field. This example illustrates a scenario in which a remote user on a dialup connection (example.net) sends mail to a border MTA (example.com) using SMTP authentication to prove identity. The dialup provider has been explicitly authorized to relay mail as "example.com" resulting in passes by the SPF and SenderID checks.
B.5. Service Provided, Several Authentications Done, Different MTAs
A message that was relayed inbound by two different MTAs that conform to this specification and applied multiple message authentication checks: Authentication-Results: example.com; sender-id=hardfail header.from=example.com; dkim=pass (good signature) header.i=sender@example.com Received: from mail-router.example.com (mail-router.example.com [192.0.2.1]) by auth-checker.example.com (8.11.6/8.11.6) with ESMTP id i7PK0sH7021929; Fri, Feb 15 2002 17:19:22 -0800 Authentication-Results: example.com; auth=pass (cram-md5) smtp.auth=sender@example.com; spf=hardfail smtp.mailfrom=example.com Received: from dialup-1-2-3-4.example.net (dialup-1-2-3-4.example.net [192.0.2.200]) by mail-router.example.com (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 DKIM-Signature: v=1; a=rsa-sha256; s=gatsby; d=example.com; i=sender@example.com; t=1188964191; c=simple/simple; h=From:Date:To:Message-Id:Subject; bh=sEuZGD/pSr7ANysbY3jtdaQ3Xv9xPQtS0m70; b=EToRSuvUfQVP3Bkz ... rTB0t0gYnBVCM= From: sender@example.com Date: Fri, Feb 15 2002 16:54:30 -0800 To: receiver@example.com Message-Id: <12345.abc@example.com> Subject: here's a sample Hello! Goodbye! Example 5: Headers reporting results from multiple MTAs The "Authentication-Results" header field is present, indicating conformance to this specification. Once again, the authserv-id used is the recipient's DNS domain name. The header field is present twice because two different MTAs in the chain of delivery did authentication tests. The first, "mail-router.example.com" reports that [AUTH] and [SPF] were both used, and [AUTH] passed but [SPF] failed. In the [AUTH] case, additional data is provided in the comment field, which the MUA can choose to render if desired.
The second MTA, "auth-checker.example.com", reports that it did a [SENDERID] test (which failed) and a [DKIM] test (which passed). Again, additional data about one of the tests is provided as a comment, which the MUA may choose to render. Since different hosts did the two sets of authentication checks, the header fields cannot be consolidated in this example. This example illustrates more typical transmission of mail into "example.com" from a user on a dialup connection "example.net". The user appears to be legitimate as he/she had a valid password allowing authentication at the border MTA using [AUTH]. The [SPF] and [SENDERID] tests failed since "example.com" has not granted "example.net" authority to relay mail on its behalf. However, the [DKIM] test passed because the sending user had a private key matching one of "example.com"'s published public keys and used it to sign the message.
B.6. Service Provided, Multi-Tiered Authentication Done
A message that had authentication done at various stages, one of which was outside the receiving ADMD: Authentication-Results: example.com; dkim=pass (good signature) header.i=@mail-router.example.net; dkim=fail (bad signature) header.i=@newyork.example.com Received: from mail-router.example.net (mail-router.example.net [192.0.2.250]) by chicago.example.com (8.11.6/8.11.6) for <recipient@chicago.example.com> with ESMTP id i7PK0sH7021929; Fri, Feb 15 2002 17:19:22 -0800 DKIM-Signature: v=1; a=rsa-sha256; s=furble; d=mail-router.example.net; t=1188964198; c=relaxed/simple; h=From:Date:To:Message-Id:Subject:Authentication-Results; bh=ftA9J6GtX8OpwUECzHnCkRzKw1uk6FNiLfJl5Nmv49E=; b=oINEO8hgn/gnunsg ... 9n9ODSNFSDij3= Authentication-Results: example.net; dkim=pass (good signature) header.i=@newyork.example.com Received: from smtp.newyork.example.com (smtp.newyork.example.com [192.0.2.220]) by mail-router.example.net (8.11.6/8.11.6) with ESMTP id g1G0r1kA003489; Fri, Feb 15 2002 17:19:07 -0800 DKIM-Signature: v=1; a=rsa-sha256; s=gatsby; d=newyork.example.com; t=1188964191; c=simple/simple; h=From:Date:To:Message-Id:Subject; bh=sEu28nfs9fuZGD/pSr7ANysbY3jtdaQ3Xv9xPQtS0m7=; b=EToRSuvUfQVP3Bkz ... rTB0t0gYnBVCM= From: sender@newyork.example.com Date: Fri, Feb 15 2002 16:54:30 -0800 To: meetings@example.net Message-Id: <12345.abc@newyork.example.com> Subject: here's a sample Example 6: Headers reporting results from multiple MTAs in different ADMDs In this example we see multi-tiered authentication with an extended trust boundary. The message was sent from someone at example.com's New York office (newyork.example.com) to a mailing list managed at an intermediary. The message was signed at the origin using [DKIM].
The message was sent to a mailing list service provider called example.net, which is used by example.com. There, meetings@example.net is expanded to a long list of recipients, one of that is at the Chicago office. In this example, we will assume that the trust boundary for chicago.example.com includes the mailing list server at example.net. The mailing list server there first authenticated the message and affixed an Authentication-Results header field indicating such using its DNS domain name for the authserv-id. It then altered the message by affixing some footer text to the body, including some administrivia such as unsubscription instructions. Finally, the mailing list server affixes a second [DKIM] signature and begins distribution of the message. The border MTA for chicago.example.com explicitly trusts results from mail-router.example.net so that header field is not removed. It performs evaluation of both signatures and determines that the first (most recent) is a "pass" but, because of the aforementioned modifications, the second is a "fail". However, the first signature included the Authentication-Results header added at mail- router.example.net that validated the second signature. Thus, indirectly, it can be determined that the authentications claimed by both signatures are indeed valid.
Appendix C. Operational Considerations about Message Authentication
This proposal is predicated on the idea that authentication (and presumably in the future, reputation) work is typically done by border MTAs rather than MUAs or intermediate MTAs; the latter merely make use of the results determined by the former. Certainly this is not mandatory for participation in electronic mail or message authentication, but the work of this proposal and its deployment to date is based on that model. The assumption satisfies several common ADMD requirements: 1. Service operators prefer to resolve the handling of problem messages as close to the border of the ADMD as possible. This enables, for example, rejections of messages at the SMTP level rather than generating a DSN internally. Thus, doing any of the authentication or reputation work exclusively at the MUA or intermediate MTA renders this desire unattainable. 2. Border MTAs are more likely to have direct access to external sources of authentication or reputation information since modern MUAs are more likely to be heavily firewalled. Thus, some MUAs might not even be able to complete the task of performing authentication or reputation evaluations without complex proxy configurations or similar burdens. 3. MUAs rely upon the upstream MTAs within their trust boundaries to make correct (as much as that is possible) evaluations about the message's envelope, header and content. Thus, MUAs don't need to know how to do the work that upstream MTAs do; they only need the results of that work. 4. Evaluations about the quality of a message, from simple token matching (e.g., a list of preferred DNS domains) to cryptanalysis (e.g., public/private key work), are at least a little bit expensive and thus should be minimized. To that end, performing those tests at the border MTA is far preferred to doing that work at each MUA that handles a message. If an ADMD's environment adheres to common messaging protocols, a reputation query or an authentication check performed by a border MTA would return the same result as the same query performed by an MUA. By contrast, in an environment where the MUA does the work, a message arriving for multiple recipients would thus cause authentication or reputation evaluation to be done more than once for the same message (i.e., at each MUA) causing needless amplification of resource use and creating a possible denial-of-service attack vector.
5. Minimizing change is good. As new authentication and reputation
methods emerge, the list of methods supported by this header
field would presumably be extended. If MUAs simply consume the
contents of this header field rather than actually attempting to
do authentication and/or reputation work, then MUAs only need to
learn to parse this header field once; emergence of new methods
requires only a configuration change at the MUAs and software
changes at the MTAs (which are presumably fewer in number). When
choosing to implement these functions in MTAs vs MUAs, the issues
of individual flexibility, infrastructure inertia and scale of
effort must be considered. It is typically easier to change a
single MUA than an MTA because the modification affects fewer
users and can be pursued with less care. However, changing many
MUAs is more effort than changing a smaller number of MTAs.
6. For decisions affecting message delivery and display, assessment
based on authentication and reputation is best performed close to
the time of message transit, as a message makes its journey
toward a user's inbox, not afterwards. DKIM keys and IP address
reputations, etc., can change over time or even become invalid,
and users can take a long time to read a message once delivered.
The value of this work thus degrades, perhaps quickly, once the
delivery process has completed. This seriously diminishes the
value of this work when done other than at MTAs.
Many operational choices are possible within an ADMD, including the
venue for performing authentication and/or reputation assessment.
The current specification does not dictate any of those choices.
Rather, it facilitates those cases in which information produced by
one stage of analysis needs to be transported with the message to the
next stage.
Acknowledgements
The author wishes to acknowledge the following for their review and constructive criticism of this proposal: Eric Allman, Mark Delany, Victor Duchovni, Frank Ellermann, Jim Fenton, Philip Guenther, Tony Hansen, Paul Hoffman, Scott Kitterman, Eliot Lear, John Levine, Miles Libbey, Charles Lindsey, Alexey Melnikov, Douglas Otis, Juan Altmayer Pizzorno, Michael Thomas, and Kazu Yamamoto. Special thanks to Dave Crocker and S. Moonesamy for their logistical support, and feedback on and contributions to the numerous proposed edits throughout the lifetime of this work.Author's Address
Murray S. Kucherawy Sendmail, Inc. 6475 Christie Ave., Suite 350 Emeryville, CA 94608 US Phone: +1 510 594 5400 EMail: msk+ietf@sendmail.com