Commit | Line | Data |
---|---|---|
22e6f294 JH |
1 | |
2 | ||
3 | ||
4 | ||
5 | DANE V. Dukhovni | |
6 | Internet-Draft Two Sigma | |
7 | Intended status: Standards Track W. Hardaker | |
8 | Expires: November 26, 2014 Parsons | |
9 | May 25, 2014 | |
10 | ||
11 | ||
12 | SMTP security via opportunistic DANE TLS | |
13 | draft-ietf-dane-smtp-with-dane-10 | |
14 | ||
15 | Abstract | |
16 | ||
17 | This memo describes a downgrade-resistant protocol for SMTP transport | |
18 | security between Mail Transfer Agents (MTAs) based on the DNS-Based | |
19 | Authentication of Named Entities (DANE) TLSA DNS record. Adoption of | |
20 | this protocol enables an incremental transition of the Internet email | |
21 | backbone to one using encrypted and authenticated Transport Layer | |
22 | Security (TLS). | |
23 | ||
24 | Status of This Memo | |
25 | ||
26 | This Internet-Draft is submitted in full conformance with the | |
27 | provisions of BCP 78 and BCP 79. | |
28 | ||
29 | Internet-Drafts are working documents of the Internet Engineering | |
30 | Task Force (IETF). Note that other groups may also distribute | |
31 | working documents as Internet-Drafts. The list of current Internet- | |
32 | Drafts is at http://datatracker.ietf.org/drafts/current/. | |
33 | ||
34 | Internet-Drafts are draft documents valid for a maximum of six months | |
35 | and may be updated, replaced, or obsoleted by other documents at any | |
36 | time. It is inappropriate to use Internet-Drafts as reference | |
37 | material or to cite them other than as "work in progress." | |
38 | ||
39 | This Internet-Draft will expire on November 26, 2014. | |
40 | ||
41 | Copyright Notice | |
42 | ||
43 | Copyright (c) 2014 IETF Trust and the persons identified as the | |
44 | document authors. All rights reserved. | |
45 | ||
46 | This document is subject to BCP 78 and the IETF Trust's Legal | |
47 | Provisions Relating to IETF Documents | |
48 | (http://trustee.ietf.org/license-info) in effect on the date of | |
49 | publication of this document. Please review these documents | |
50 | carefully, as they describe your rights and restrictions with respect | |
51 | to this document. Code Components extracted from this document must | |
52 | include Simplified BSD License text as described in Section 4.e of | |
53 | ||
54 | ||
55 | ||
56 | Dukhovni & Hardaker Expires November 26, 2014 [Page 1] | |
57 | \f | |
58 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
59 | ||
60 | ||
61 | the Trust Legal Provisions and are provided without warranty as | |
62 | described in the Simplified BSD License. | |
63 | ||
64 | Table of Contents | |
65 | ||
66 | 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 | |
67 | 1.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 3 | |
68 | 1.2. Background . . . . . . . . . . . . . . . . . . . . . . . 5 | |
69 | 1.3. SMTP channel security . . . . . . . . . . . . . . . . . . 6 | |
70 | 1.3.1. STARTTLS downgrade attack . . . . . . . . . . . . . . 6 | |
71 | 1.3.2. Insecure server name without DNSSEC . . . . . . . . . 7 | |
72 | 1.3.3. Sender policy does not scale . . . . . . . . . . . . 7 | |
73 | 1.3.4. Too many certification authorities . . . . . . . . . 8 | |
74 | 2. Identifying applicable TLSA records . . . . . . . . . . . . . 8 | |
75 | 2.1. DNS considerations . . . . . . . . . . . . . . . . . . . 8 | |
76 | 2.1.1. DNS errors, bogus and indeterminate responses . . . . 8 | |
77 | 2.1.2. DNS error handling . . . . . . . . . . . . . . . . . 11 | |
78 | 2.1.3. Stub resolver considerations . . . . . . . . . . . . 11 | |
79 | 2.2. TLS discovery . . . . . . . . . . . . . . . . . . . . . . 12 | |
80 | 2.2.1. MX resolution . . . . . . . . . . . . . . . . . . . . 13 | |
81 | 2.2.2. Non-MX destinations . . . . . . . . . . . . . . . . . 15 | |
82 | 2.2.3. TLSA record lookup . . . . . . . . . . . . . . . . . 17 | |
83 | 3. DANE authentication . . . . . . . . . . . . . . . . . . . . . 19 | |
84 | 3.1. TLSA certificate usages . . . . . . . . . . . . . . . . . 19 | |
85 | 3.1.1. Certificate usage DANE-EE(3) . . . . . . . . . . . . 20 | |
86 | 3.1.2. Certificate usage DANE-TA(2) . . . . . . . . . . . . 21 | |
87 | 3.1.3. Certificate usages PKIX-TA(0) and PKIX-EE(1) . . . . 22 | |
88 | 3.2. Certificate matching . . . . . . . . . . . . . . . . . . 23 | |
89 | 3.2.1. DANE-EE(3) name checks . . . . . . . . . . . . . . . 23 | |
90 | 3.2.2. DANE-TA(2) name checks . . . . . . . . . . . . . . . 23 | |
91 | 3.2.3. Reference identifier matching . . . . . . . . . . . . 24 | |
92 | 4. Server key management . . . . . . . . . . . . . . . . . . . . 25 | |
93 | 5. Digest algorithm agility . . . . . . . . . . . . . . . . . . 26 | |
94 | 6. Mandatory TLS Security . . . . . . . . . . . . . . . . . . . 27 | |
95 | 7. Note on DANE for Message User Agents . . . . . . . . . . . . 28 | |
96 | 8. Interoperability considerations . . . . . . . . . . . . . . . 29 | |
97 | 8.1. SNI support . . . . . . . . . . . . . . . . . . . . . . . 29 | |
98 | 8.2. Anonymous TLS cipher suites . . . . . . . . . . . . . . . 29 | |
99 | 9. Operational Considerations . . . . . . . . . . . . . . . . . 30 | |
100 | 9.1. Client Operational Considerations . . . . . . . . . . . . 30 | |
101 | 9.2. Publisher Operational Considerations . . . . . . . . . . 30 | |
102 | 10. Security Considerations . . . . . . . . . . . . . . . . . . . 31 | |
103 | 11. IANA considerations . . . . . . . . . . . . . . . . . . . . . 31 | |
104 | 12. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 31 | |
105 | 13. References . . . . . . . . . . . . . . . . . . . . . . . . . 32 | |
106 | 13.1. Normative References . . . . . . . . . . . . . . . . . . 32 | |
107 | 13.2. Informative References . . . . . . . . . . . . . . . . . 33 | |
108 | Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 33 | |
109 | ||
110 | ||
111 | ||
112 | Dukhovni & Hardaker Expires November 26, 2014 [Page 2] | |
113 | \f | |
114 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
115 | ||
116 | ||
117 | 1. Introduction | |
118 | ||
119 | This memo specifies a new connection security model for Message | |
120 | Transfer Agents (MTAs). This model is motivated by key features of | |
121 | inter-domain SMTP delivery, in particular the fact that the | |
122 | destination server is selected indirectly via DNS Mail Exchange (MX) | |
123 | records and that neither email addresses nor MX hostnames signal a | |
124 | requirement for either secure or cleartext transport. Therefore, | |
125 | aside from a few manually configured exceptions, SMTP transport | |
126 | security is of necessity opportunistic. | |
127 | ||
128 | This specification uses the presence of DANE TLSA records to securely | |
129 | signal TLS support and to publish the means by which SMTP clients can | |
130 | successfully authenticate legitimate SMTP servers. This becomes | |
131 | "opportunistic DANE TLS" and is resistant to downgrade and MITM | |
132 | attacks. It enables an incremental transition of the email backbone | |
133 | to authenticated TLS delivery, with increased global protection as | |
134 | adoption increases. | |
135 | ||
136 | With opportunistic DANE TLS, traffic from SMTP clients to domains | |
137 | that publish "usable" DANE TLSA records in accordance with this memo | |
138 | is authenticated and encrypted. Traffic from legacy clients or to | |
139 | domains that do not publish TLSA records will continue to be sent in | |
140 | the same manner as before, via manually configured security, (pre- | |
141 | DANE) opportunistic TLS or just cleartext SMTP. | |
142 | ||
143 | Problems with existing use of TLS in MTA to MTA SMTP that motivate | |
144 | this specification are described in Section 1.3. The specification | |
145 | itself follows in Section 2 and Section 3 which describe respectively | |
146 | how to locate and use DANE TLSA records with SMTP. In Section 6, we | |
147 | discuss application of DANE TLS to destinations for which channel | |
148 | integrity and confidentiality are mandatory. In Section 7 we briefly | |
149 | comment on potential applicability of this specification to Message | |
150 | User Agents. | |
151 | ||
152 | 1.1. Terminology | |
153 | ||
154 | The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", | |
155 | "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and | |
156 | "OPTIONAL" in this document are to be interpreted as described in | |
157 | [RFC2119]. | |
158 | ||
159 | The following terms or concepts are used through the document: | |
160 | ||
161 | Man-in-the-middle or MITM attack: Active modification of network | |
162 | traffic by an adversary able to thereby compromise the | |
163 | confidentiality or integrity of the data. | |
164 | ||
165 | ||
166 | ||
167 | ||
168 | Dukhovni & Hardaker Expires November 26, 2014 [Page 3] | |
169 | \f | |
170 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
171 | ||
172 | ||
173 | secure, bogus, insecure, indeterminate: DNSSEC validation results, | |
174 | as defined in Section 4.3 of [RFC4035]. | |
175 | ||
176 | Validating Security-Aware Stub Resolver and Non-Validating | |
177 | Security-Aware Stub Resolver: | |
178 | Capabilities of the stub resolver in use as defined in [RFC4033]; | |
179 | note that this specification requires the use of a Security-Aware | |
180 | Stub Resolver; Security-Oblivious stub-resolvers MUST NOT be used. | |
181 | ||
182 | opportunistic DANE TLS: Best-effort use of TLS, resistant to | |
183 | downgrade attacks for destinations with DNSSEC-validated TLSA | |
184 | records. When opportunistic DANE TLS is determined to be | |
185 | unavailable, clients should fall back to opportunistic TLS below. | |
186 | Opportunistic DANE TLS requires support for DNSSEC, DANE and | |
187 | STARTTLS on the client side and STARTTLS plus a DNSSEC published | |
188 | TLSA record on the server side. | |
189 | ||
190 | (pre-DANE) opportunistic TLS: Best-effort use of TLS that is | |
191 | generally vulnerable to DNS forgery and STARTTLS downgrade | |
192 | attacks. When a TLS-encrypted communication channel is not | |
193 | available, message transmission takes place in the clear. MX | |
194 | record indirection generally precludes authentication even when | |
195 | TLS is available. | |
196 | ||
197 | reference identifier: (Special case of [RFC6125] definition). One | |
198 | of the domain names associated by the SMTP client with the | |
199 | destination SMTP server for performing name checks on the server | |
200 | certificate. When name checks are applicable, at least one of the | |
201 | reference identifiers MUST match an [RFC6125] DNS-ID (or if none | |
202 | are present the [RFC6125] CN-ID) of the server certificate (see | |
203 | Section 3.2.3). | |
204 | ||
205 | MX hostname: The RRDATA of an MX record consists of a 16 bit | |
206 | preference followed by a Mail Exchange domain name (see [RFC1035], | |
207 | Section 3.3.9). We will use the term "MX hostname" to refer to | |
208 | the latter, that is, the DNS domain name found after the | |
209 | preference value in an MX record. Thus an "MX hostname" is | |
210 | specifically a reference to a DNS domain name, rather than any | |
211 | host that bears that name. | |
212 | ||
213 | delayed delivery: Email delivery is a multi-hop store & forward | |
214 | process. When an MTA is unable forward a message that may become | |
215 | deliverable later, the message is queued and delivery is retried | |
216 | periodically. Some MTAs may be configured with a fallback next- | |
217 | hop destination that handles messages that the MTA would otherwise | |
218 | queue and retry. In these cases, messages that would otherwise | |
219 | have to be delayed, may be sent to the fallback next-hop | |
220 | destination instead. The fallback destination may itself be | |
221 | ||
222 | ||
223 | ||
224 | Dukhovni & Hardaker Expires November 26, 2014 [Page 4] | |
225 | \f | |
226 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
227 | ||
228 | ||
229 | subject to opportunistic or mandatory DANE TLS as though it were | |
230 | the original message destination. | |
231 | ||
232 | original next hop destination: The logical destination for mail | |
233 | delivery. By default this is the domain portion of the recipient | |
234 | address, but MTAs may be configured to forward mail for some or | |
235 | all recipients via designated relays. The original next hop | |
236 | destination is, respectively, either the recipient domain or the | |
237 | associated configured relay. | |
238 | ||
239 | MTA: Message Transfer Agent ([RFC5598], Section 4.3.2). | |
240 | ||
241 | MSA: Message Submission Agent ([RFC5598], Section 4.3.1). | |
242 | ||
243 | MUA: Message User Agent ([RFC5598], Section 4.2.1). | |
244 | ||
245 | RR: A DNS Resource Record | |
246 | ||
247 | RRset: A set of DNS Resource Records for a particular class, domain | |
248 | and record type. | |
249 | ||
250 | 1.2. Background | |
251 | ||
252 | The Domain Name System Security Extensions (DNSSEC) add data origin | |
253 | authentication, data integrity and data non-existence proofs to the | |
254 | Domain Name System (DNS). DNSSEC is defined in [RFC4033], [RFC4034] | |
255 | and [RFC4035]. | |
256 | ||
257 | As described in the introduction of [RFC6698], TLS authentication via | |
258 | the existing public Certification Authority (CA) PKI suffers from an | |
259 | over-abundance of trusted parties capable of issuing certificates for | |
260 | any domain of their choice. DANE leverages the DNSSEC infrastructure | |
261 | to publish trusted public keys and certificates for use with the | |
262 | Transport Layer Security (TLS) [RFC5246] protocol via a new "TLSA" | |
263 | DNS record type. With DNSSEC each domain can only vouch for the keys | |
264 | of its directly delegated sub-domains. | |
265 | ||
266 | The TLS protocol enables secure TCP communication. In the context of | |
267 | this memo, channel security is assumed to be provided by TLS. Used | |
268 | without authentication, TLS provides only privacy protection against | |
269 | eavesdropping attacks. With authentication, TLS also provides data | |
270 | integrity protection to guard against MITM attacks. | |
271 | ||
272 | ||
273 | ||
274 | ||
275 | ||
276 | ||
277 | ||
278 | ||
279 | ||
280 | Dukhovni & Hardaker Expires November 26, 2014 [Page 5] | |
281 | \f | |
282 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
283 | ||
284 | ||
285 | 1.3. SMTP channel security | |
286 | ||
287 | With HTTPS, Transport Layer Security (TLS) employs X.509 certificates | |
288 | [RFC5280] issued by one of the many Certificate Authorities (CAs) | |
289 | bundled with popular web browsers to allow users to authenticate | |
290 | their "secure" websites. Before we specify a new DANE TLS security | |
291 | model for SMTP, we will explain why a new security model is needed. | |
292 | In the process, we will explain why the familiar HTTPS security model | |
293 | is inadequate to protect inter-domain SMTP traffic. | |
294 | ||
295 | The subsections below outline four key problems with applying | |
296 | traditional PKI to SMTP that are addressed by this specification. | |
297 | Since SMTP channel security policy is not explicitly specified in | |
298 | either the recipient address or the MX record, a new signaling | |
299 | mechanism is required to indicate when channel security is possible | |
300 | and should be used. The publication of TLSA records allows server | |
301 | operators to securely signal to SMTP clients that TLS is available | |
302 | and should be used. DANE TLSA makes it possible to simultaneously | |
303 | discover which destination domains support secure delivery via TLS | |
304 | and how to verify the authenticity of the associated SMTP services, | |
305 | providing a path forward to ubiquitous SMTP channel security. | |
306 | ||
307 | 1.3.1. STARTTLS downgrade attack | |
308 | ||
309 | The Simple Mail Transfer Protocol (SMTP) [RFC5321] is a single-hop | |
310 | protocol in a multi-hop store & forward email delivery process. SMTP | |
311 | envelope recipient addresses are not transport addresses and are | |
312 | security-agnostic. Unlike the Hypertext Transfer Protocol (HTTP) and | |
313 | its corresponding secured version, HTTPS, where the use of TLS is | |
314 | signaled via the URI scheme, email recipient addresses do not | |
315 | directly signal transport security policy. Indeed, no such signaling | |
316 | could work well with SMTP since TLS encryption of SMTP protects email | |
317 | traffic on a hop-by-hop basis while email addresses could only | |
318 | express end-to-end policy. | |
319 | ||
320 | With no mechanism available to signal transport security policy, SMTP | |
321 | relays employ a best-effort "opportunistic" security model for TLS. | |
322 | A single SMTP server TCP listening endpoint can serve both TLS and | |
323 | non-TLS clients; the use of TLS is negotiated via the SMTP STARTTLS | |
324 | command ([RFC3207]). The server signals TLS support to the client | |
325 | over a cleartext SMTP connection, and, if the client also supports | |
326 | TLS, it may negotiate a TLS encrypted channel to use for email | |
327 | transmission. The server's indication of TLS support can be easily | |
328 | suppressed by an MITM attacker. Thus pre-DANE SMTP TLS security can | |
329 | be subverted by simply downgrading a connection to cleartext. No TLS | |
330 | security feature, such as the use of PKIX, can prevent this. The | |
331 | attacker can simply disable TLS. | |
332 | ||
333 | ||
334 | ||
335 | ||
336 | Dukhovni & Hardaker Expires November 26, 2014 [Page 6] | |
337 | \f | |
338 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
339 | ||
340 | ||
341 | 1.3.2. Insecure server name without DNSSEC | |
342 | ||
343 | With SMTP, DNS Mail Exchange (MX) records abstract the next-hop | |
344 | transport endpoint and allow administrators to specify a set of | |
345 | target servers to which SMTP traffic should be directed for a given | |
346 | domain. | |
347 | ||
348 | A PKIX TLS client is vulnerable to MITM attacks unless it verifies | |
349 | that the server's certificate binds the public key to a name that | |
350 | matches one of the client's reference identifiers. A natural choice | |
351 | of reference identifier is the server's domain name. However, with | |
352 | SMTP, server names are obtained indirectly via MX records. Without | |
353 | DNSSEC, the MX lookup is vulnerable to MITM and DNS cache poisoning | |
354 | attacks. Active attackers can forge DNS replies with fake MX records | |
355 | and can redirect email to servers with names of their choice. | |
356 | Therefore, secure verification of SMTP TLS certificates matching the | |
357 | server name is not possible without DNSSEC. | |
358 | ||
359 | One might try to harden TLS for SMTP against DNS attacks by using the | |
360 | envelope recipient domain as a reference identifier and requiring | |
361 | each SMTP server to possess a trusted certificate for the envelope | |
362 | recipient domain rather than the MX hostname. Unfortunately, this is | |
363 | impractical as email for many domains is handled by third parties | |
364 | that are not in a position to obtain certificates for all the domains | |
365 | they serve. Deployment of the Server Name Indication (SNI) extension | |
366 | to TLS (see [RFC6066] Section 3) is no panacea, since SNI key | |
367 | management is operationally challenging except when the email service | |
368 | provider is also the domain's registrar and its certificate issuer; | |
369 | this is rarely the case for email. | |
370 | ||
371 | Since the recipient domain name cannot be used as the SMTP server | |
372 | reference identifier, and neither can the MX hostname without DNSSEC, | |
373 | large-scale deployment of authenticated TLS for SMTP requires that | |
374 | the DNS be secure. | |
375 | ||
376 | Since SMTP security depends critically on DNSSEC, it is important to | |
377 | point out that consequently SMTP with DANE is the most conservative | |
378 | possible trust model. It trusts only what must be trusted and no | |
379 | more. Adding any other trusted actors to the mix can only reduce | |
380 | SMTP security. A sender may choose to further harden DNSSEC for | |
381 | selected high-value receiving domains, by configuring explicit trust | |
382 | anchors for those domains instead of relying on the chain of trust | |
383 | from the root domain. Detailed discussion of DNSSEC security | |
384 | practices is out of scope for this document. | |
385 | ||
386 | 1.3.3. Sender policy does not scale | |
387 | ||
388 | ||
389 | ||
390 | ||
391 | ||
392 | Dukhovni & Hardaker Expires November 26, 2014 [Page 7] | |
393 | \f | |
394 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
395 | ||
396 | ||
397 | Sending systems are in some cases explicitly configured to use TLS | |
398 | for mail sent to selected peer domains. This requires sending MTAs | |
399 | to be configured with appropriate subject names or certificate | |
400 | content digests to expect in the presented server certificates. | |
401 | Because of the heavy administrative burden, such statically | |
402 | configured SMTP secure channels are used rarely (generally only | |
403 | between domains that make bilateral arrangements with their business | |
404 | partners). Internet email, on the other hand, requires regularly | |
405 | contacting new domains for which security configurations cannot be | |
406 | established in advance. | |
407 | ||
408 | The abstraction of the SMTP transport endpoint via DNS MX records, | |
409 | often across organization boundaries, limits the use of public CA PKI | |
410 | with SMTP to a small set of sender-configured peer domains. With | |
411 | little opportunity to use TLS authentication, sending MTAs are rarely | |
412 | configured with a comprehensive list of trusted CAs. SMTP services | |
413 | that support STARTTLS often deploy X.509 certificates that are self- | |
414 | signed or issued by a private CA. | |
415 | ||
416 | 1.3.4. Too many certification authorities | |
417 | ||
418 | Even if it were generally possible to determine a secure server name, | |
419 | the SMTP client would still need to verify that the server's | |
420 | certificate chain is issued by a trusted Certification Authority (a | |
421 | trust anchor). MTAs are not interactive applications where a human | |
422 | operator can make a decision (wisely or otherwise) to selectively | |
423 | disable TLS security policy when certificate chain verification | |
424 | fails. With no user to "click OK", the MTAs list of public CA trust | |
425 | anchors would need to be comprehensive in order to avoid bouncing | |
426 | mail addressed to sites that employ unknown Certification | |
427 | Authorities. | |
428 | ||
429 | On the other hand, each trusted CA can issue certificates for any | |
430 | domain. If even one of the configured CAs is compromised or operated | |
431 | by an adversary, it can subvert TLS security for all destinations. | |
432 | Any set of CAs is simultaneously both overly inclusive and not | |
433 | inclusive enough. | |
434 | ||
435 | 2. Identifying applicable TLSA records | |
436 | ||
437 | 2.1. DNS considerations | |
438 | ||
439 | 2.1.1. DNS errors, bogus and indeterminate responses | |
440 | ||
441 | ||
442 | ||
443 | ||
444 | ||
445 | ||
446 | ||
447 | ||
448 | Dukhovni & Hardaker Expires November 26, 2014 [Page 8] | |
449 | \f | |
450 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
451 | ||
452 | ||
453 | An SMTP client that implements opportunistic DANE TLS per this | |
454 | specification depends critically on the integrity of DNSSEC lookups, | |
455 | as discussed in Section 1.3. This section lists the DNS resolver | |
456 | requirements needed to avoid downgrade attacks when using | |
457 | opportunistic DANE TLS. | |
458 | ||
459 | A DNS lookup may signal an error or return a definitive answer. A | |
460 | security-aware resolver must be used for this specification. | |
461 | Security-aware resolvers will indicate the security status of a DNS | |
462 | RRset with one of four possible values defined in Section 4.3 of | |
463 | [RFC4035]: "secure", "insecure", "bogus" and "indeterminate". In | |
464 | [RFC4035] the meaning of the "indeterminate" security status is: | |
465 | ||
466 | An RRset for which the resolver is not able to determine whether | |
467 | the RRset should be signed, as the resolver is not able to obtain | |
468 | the necessary DNSSEC RRs. This can occur when the security-aware | |
469 | resolver is not able to contact security-aware name servers for | |
470 | the relevant zones. | |
471 | ||
472 | Note, the "indeterminate" security status has a conflicting | |
473 | definition in section 5 of [RFC4033]. | |
474 | ||
475 | There is no trust anchor that would indicate that a specific | |
476 | portion of the tree is secure. | |
477 | ||
478 | SMTP clients following this specification SHOULD NOT distinguish | |
479 | between "insecure" and "indeterminate" in the [RFC4033] sense. Both | |
480 | "insecure" and RFC4033 "indeterminate" are handled identically: in | |
481 | either case unvalidated data for the query domain is all that is and | |
482 | can be available, and authentication using the data is impossible. | |
483 | In what follows, when we say "insecure", we include also DNS results | |
484 | for domains that lie in a portion of the DNS tree for which there is | |
485 | no applicable trust anchor. With the DNS root zone signed, we expect | |
486 | that validating resolvers used by Internet-facing MTAs will be | |
487 | configured with trust anchor data for the root zone. Therefore, | |
488 | RFC4033-style "indeterminate" domains should be rare in practice. | |
489 | From here on, when we say "indeterminate", it is exclusively in the | |
490 | sense of [RFC4035]. | |
491 | ||
492 | As noted in section 4.3 of [RFC4035], a security-aware DNS resolver | |
493 | MUST be able to determine whether a given non-error DNS response is | |
494 | "secure", "insecure", "bogus" or "indeterminate". It is expected | |
495 | that most security-aware stub resolvers will not signal an | |
496 | "indeterminate" security status in the RFC4035-sense to the | |
497 | application, and will signal a "bogus" or error result instead. If a | |
498 | resolver does signal an RFC4035 "indeterminate" security status, this | |
499 | MUST be treated by the SMTP client as though a "bogus" or error | |
500 | result had been returned. | |
501 | ||
502 | ||
503 | ||
504 | Dukhovni & Hardaker Expires November 26, 2014 [Page 9] | |
505 | \f | |
506 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
507 | ||
508 | ||
509 | An MTA making use of a non-validating security-aware stub resolver | |
510 | MAY use the stub resolver's ability, if available, to signal DNSSEC | |
511 | validation status based on information the stub resolver has learned | |
512 | from an upstream validating recursive resolver. In accordance with | |
513 | section 4.9.3 of [RFC4035]: | |
514 | ||
515 | ... a security-aware stub resolver MUST NOT place any reliance on | |
516 | signature validation allegedly performed on its behalf, except | |
517 | when the security-aware stub resolver obtained the data in question | |
518 | from a trusted security-aware recursive name server via a secure | |
519 | channel. | |
520 | ||
521 | To avoid much repetition in the text below, we will pause to explain | |
522 | the handling of "bogus" or "indeterminate" DNSSEC query responses. | |
523 | These are not necessarily the result of a malicious actor; they can, | |
524 | for example, occur when network packets are corrupted or lost in | |
525 | transit. Therefore, "bogus" or "indeterminate" replies are equated | |
526 | in this memo with lookup failure. | |
527 | ||
528 | There is an important non-failure condition we need to highlight in | |
529 | addition to the obvious case of the DNS client obtaining a non-empty | |
530 | "secure" or "insecure" RRset of the requested type. Namely, it is | |
531 | not an error when either "secure" or "insecure" non-existence is | |
532 | determined for the requested data. When a DNSSEC response with a | |
533 | validation status that is either "secure" or "insecure" reports | |
534 | either no records of the requested type or non-existence of the query | |
535 | domain, the response is not a DNS error condition. The DNS client | |
536 | has not been left without an answer; it has learned that records of | |
537 | the requested type do not exist. | |
538 | ||
539 | Security-aware stub resolvers will, of course, also signal DNS lookup | |
540 | errors in other cases, for example when processing a "ServFail" | |
541 | RCODE, which will not have an associated DNSSEC status. All lookup | |
542 | errors are treated the same way by this specification, regardless of | |
543 | whether they are from a "bogus" or "indeterminate" DNSSEC status or | |
544 | from a more generic DNS error: the information that was requested | |
545 | cannot be obtained by the security-aware resolver at this time. A | |
546 | lookup error is thus a failure to obtain the relevant RRset if it | |
547 | exists, or to determine that no such RRset exists when it does not. | |
548 | ||
549 | In contrast to a "bogus" or an "indeterminate" response, an | |
550 | "insecure" DNSSEC response is not an error, rather it indicates that | |
551 | the target DNS zone is either securely opted out of DNSSEC validation | |
552 | or is not connected with the DNSSEC trust anchors being used. | |
553 | Insecure results will leave the SMTP client with degraded channel | |
554 | security, but do not stand in the way of message delivery. See | |
555 | section Section 2.2 for further details. | |
556 | ||
557 | ||
558 | ||
559 | ||
560 | Dukhovni & Hardaker Expires November 26, 2014 [Page 10] | |
561 | \f | |
562 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
563 | ||
564 | ||
565 | 2.1.2. DNS error handling | |
566 | ||
567 | When a DNS lookup failure (error or "bogus" or "indeterminate" as | |
568 | defined above) prevents an SMTP client from determining which SMTP | |
569 | server or servers it should connect to, message delivery MUST be | |
570 | delayed. This naturally includes, for example, the case when a | |
571 | "bogus" or "indeterminate" response is encountered during MX | |
572 | resolution. When multiple MX hostnames are obtained from a | |
573 | successful MX lookup, but a later DNS lookup failure prevents network | |
574 | address resolution for a given MX hostname, delivery may proceed via | |
575 | any remaining MX hosts. | |
576 | ||
577 | When a particular SMTP server is securely identified as the delivery | |
578 | destination, a set of DNS lookups (Section 2.2) MUST be performed to | |
579 | locate any related TLSA records. If any DNS queries used to locate | |
580 | TLSA records fail (be it due to "bogus" or "indeterminate" records, | |
581 | timeouts, malformed replies, ServFails, etc.), then the SMTP client | |
582 | MUST treat that server as unreachable and MUST NOT deliver the | |
583 | message via that server. If no servers are reachable, delivery is | |
584 | delayed. | |
585 | ||
586 | In what follows, we will only describe what happens when all relevant | |
587 | DNS queries succeed. If any DNS failure occurs, the SMTP client MUST | |
588 | behave as described in this section, by skipping the problem SMTP | |
589 | server, or the problem destination. Queries for candidate TLSA | |
590 | records are explicitly part of "all relevant DNS queries" and SMTP | |
591 | clients MUST NOT continue to connect to an SMTP server or destination | |
592 | whose TLSA record lookup fails. | |
593 | ||
594 | 2.1.3. Stub resolver considerations | |
595 | ||
596 | A note about DNAME aliases: a query for a domain name whose ancestor | |
597 | domain is a DNAME alias returns the DNAME RR for the ancestor domain, | |
598 | along with a CNAME that maps the query domain to the corresponding | |
599 | sub-domain of the target domain of the DNAME alias [RFC6672]. | |
600 | Therefore, whenever we speak of CNAME aliases, we implicitly allow | |
601 | for the possibility that the alias in question is the result of an | |
602 | ancestor domain DNAME record. Consequently, no explicit support for | |
603 | DNAME records is needed in SMTP software, it is sufficient to process | |
604 | the resulting CNAME aliases. DNAME records only require special | |
605 | processing in the validating stub-resolver library that checks the | |
606 | integrity of the combined DNAME + CNAME reply. When DNSSEC | |
607 | validation is handled by a local caching resolver, rather than the | |
608 | MTA itself, even that part of the DNAME support logic is outside the | |
609 | MTA. | |
610 | ||
611 | When a stub resolver returns a response containing a CNAME alias that | |
612 | does not also contain the corresponding query results for the target | |
613 | ||
614 | ||
615 | ||
616 | Dukhovni & Hardaker Expires November 26, 2014 [Page 11] | |
617 | \f | |
618 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
619 | ||
620 | ||
621 | of the alias, the SMTP client will need to repeat the query at the | |
622 | target of the alias, and should do so recursively up to some | |
623 | configured or implementation-dependent recursion limit. If at any | |
624 | stage of CNAME expansion an error is detected, the lookup of the | |
625 | original requested records MUST be considered to have failed. | |
626 | ||
627 | Whether a chain of CNAME records was returned in a single stub | |
628 | resolver response or via explicit recursion by the SMTP client, if at | |
629 | any stage of recursive expansion an "insecure" CNAME record is | |
630 | encountered, then it and all subsequent results (in particular, the | |
631 | final result) MUST be considered "insecure" regardless of whether any | |
632 | earlier CNAME records leading to the "insecure" record were "secure". | |
633 | ||
634 | Note, a security-aware non-validating stub resolver may return to the | |
635 | SMTP client an "insecure" reply received from a validating recursive | |
636 | resolver that contains a CNAME record along with additional answers | |
637 | recursively obtained starting at the target of the CNAME. In this | |
638 | all that one can say is that some record in the set of records | |
639 | returned is "insecure", but it is possible that the initial CNAME | |
640 | record and a subset of the subsequent records are "secure". | |
641 | ||
642 | If the SMTP client needs to determine the security status of the DNS | |
643 | zone containing the initial CNAME record, it may need to issue an a | |
644 | separate query of type "CNAME" that returns only the initial CNAME | |
645 | record. In particular in Section 2.2.2 when insecure A or AAAA | |
646 | records are found for an SMTP server via a CNAME alias, it may be | |
647 | necessary to perform an additional CNAME query to determine whether | |
648 | the DNS zone in which the alias is published is signed. | |
649 | ||
650 | 2.2. TLS discovery | |
651 | ||
652 | As noted previously (in Section 1.3.1), opportunistic TLS with SMTP | |
653 | servers that advertise TLS support via STARTTLS is subject to an MITM | |
654 | downgrade attack. Also some SMTP servers that are not, in fact, TLS | |
655 | capable erroneously advertise STARTTLS by default and clients need to | |
656 | be prepared to retry cleartext delivery after STARTTLS fails. In | |
657 | contrast, DNSSEC validated TLSA records MUST NOT be published for | |
658 | servers that do not support TLS. Clients can safely interpret their | |
659 | presence as a commitment by the server operator to implement TLS and | |
660 | STARTTLS. | |
661 | ||
662 | This memo defines four actions to be taken after the search for a | |
663 | TLSA record returns secure usable results, secure unusable results, | |
664 | insecure or no results or an error signal. The term "usable" in this | |
665 | context is in the sense of Section 4.1 of [RFC6698]. Specifically, | |
666 | if the DNS lookup for a TLSA record returns: | |
667 | ||
668 | ||
669 | ||
670 | ||
671 | ||
672 | Dukhovni & Hardaker Expires November 26, 2014 [Page 12] | |
673 | \f | |
674 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
675 | ||
676 | ||
677 | A secure TLSA RRset with at least one usable record: A connection to | |
678 | the MTA MUST be made using authenticated and encrypted TLS, using | |
679 | the techniques discussed in the rest of this document. Failure to | |
680 | establish an authenticated TLS connection MUST result in falling | |
681 | back to the next SMTP server or delayed delivery. | |
682 | ||
683 | A Secure non-empty TLSA RRset where all the records are unusable: A | |
684 | connection to the MTA MUST be made via TLS, but authentication is | |
685 | not required. Failure to establish an encrypted TLS connection | |
686 | MUST result in falling back to the next SMTP server or delayed | |
687 | delivery. | |
688 | ||
689 | An insecure TLSA RRset or DNSSEC validated proof-of-non-existent TLSA | |
690 | records: | |
691 | A connection to the MTA SHOULD be made using (pre-DANE) | |
692 | opportunistic TLS, this includes using cleartext delivery when the | |
693 | remote SMTP server does not appear to support TLS. The MTA MAY | |
694 | retry in cleartext when delivery via TLS fails either during the | |
695 | handshake or even during data transfer. | |
696 | ||
697 | Any lookup error: Lookup errors, including "bogus" and | |
698 | "indeterminate", as explained in Section 2.1.1 MUST result in | |
699 | falling back to the next SMTP server or delayed delivery. | |
700 | ||
701 | An SMTP client MAY be configured to require DANE verified delivery | |
702 | for some destinations. We will call such a configuration "mandatory | |
703 | DANE TLS". With mandatory DANE TLS, delivery proceeds only when | |
704 | "secure" TLSA records are used to establish an encrypted and | |
705 | authenticated TLS channel with the SMTP server. | |
706 | ||
707 | When the original next-hop destination is an address literal, rather | |
708 | than a DNS domain, DANE TLS does not apply. Delivery proceeds using | |
709 | any relevant security policy configured by the MTA administrator. | |
710 | Similarly, when an MX RRset incorrectly lists a network address in | |
711 | lieu of an MX hostname, if the MTA chooses to connect to the network | |
712 | address DANE TLSA does not apply for such a connection. | |
713 | ||
714 | In the subsections that follow we explain how to locate the SMTP | |
715 | servers and the associated TLSA records for a given next-hop | |
716 | destination domain. We also explain which name or names are to be | |
717 | used in identity checks of the SMTP server certificate. | |
718 | ||
719 | 2.2.1. MX resolution | |
720 | ||
721 | In this section we consider next-hop domains that are subject to MX | |
722 | resolution and have MX records. The TLSA records and the associated | |
723 | base domain are derived separately for each MX hostname that is used | |
724 | to attempt message delivery. DANE TLS can authenticate message | |
725 | ||
726 | ||
727 | ||
728 | Dukhovni & Hardaker Expires November 26, 2014 [Page 13] | |
729 | \f | |
730 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
731 | ||
732 | ||
733 | delivery to the intended next-hop domain only when the MX records are | |
734 | obtained securely via a DNSSEC validated lookup. | |
735 | ||
736 | MX records MUST be sorted by preference; an MX hostname with a worse | |
737 | (numerically higher) MX preference that has TLSA records MUST NOT | |
738 | preempt an MX hostname with a better (numerically lower) preference | |
739 | that has no TLSA records. In other words, prevention of delivery | |
740 | loops by obeying MX preferences MUST take precedence over channel | |
741 | security considerations. Even with two equal-preference MX records, | |
742 | an MTA is not obligated to choose the MX hostname that offers more | |
743 | security. Domains that want secure inbound mail delivery need to | |
744 | ensure that all their SMTP servers and MX records are configured | |
745 | accordingly. | |
746 | ||
747 | In the language of [RFC5321] Section 5.1, the original next-hop | |
748 | domain is the "initial name". If the MX lookup of the initial name | |
749 | results in a CNAME alias, the MTA replaces the initial name with the | |
750 | resulting name and performs a new lookup with the new name. MTAs | |
751 | typically support recursion in CNAME expansion, so this replacement | |
752 | is performed repeatedly until the ultimate non-CNAME domain is found. | |
753 | ||
754 | If the MX RRset (or any CNAME leading to it) is "insecure" (see | |
755 | Section 2.1.1), DANE TLS need not apply, and delivery MAY proceed via | |
756 | pre-DANE opportunistic TLS. That said, the protocol in this memo is | |
757 | an "opportunistic security" protocol, meaning that it strives to | |
758 | communicate with each peer as securely as possible, while maintaining | |
759 | broad interoperability. Therefore, the SMTP client MAY proceed to | |
760 | use DANE TLS (as described in Section 2.2.2 below) even with MX hosts | |
761 | obtained via an "insecure" MX RRset. For example, when a hosting | |
762 | provider has a signed DNS zone and publishes TLSA records for its | |
763 | SMTP servers, hosted domains that are not signed may still benefit | |
764 | from the provider's TLSA records. Deliveries via the provider's SMTP | |
765 | servers will not be subject to active attacks when sending SMTP | |
766 | clients elect to make use of the provider's TLSA records. | |
767 | ||
768 | When the MX records are not (DNSSEC) signed, an active attacker can | |
769 | redirect SMTP clients to MX hosts of his choice. Such redirection is | |
770 | tamper-evident when SMTP servers found via "insecure" MX records are | |
771 | recorded as the next-hop relay in the MTA delivery logs in their | |
772 | original (rather than CNAME expanded) form. Sending MTAs SHOULD log | |
773 | unexpanded MX hostnames when these result from insecure MX lookups. | |
774 | Any successful authentication via an insecurely determined MX host | |
775 | MUST NOT be misrepresented in the mail logs as secure delivery to the | |
776 | intended next-hop domain. When DANE TLS is mandatory (Section 6) for | |
777 | a given destination, delivery MUST be delayed when the MX RRset is | |
778 | not "secure". | |
779 | ||
780 | ||
781 | ||
782 | ||
783 | ||
784 | Dukhovni & Hardaker Expires November 26, 2014 [Page 14] | |
785 | \f | |
786 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
787 | ||
788 | ||
789 | Otherwise, assuming no DNS errors (Section 2.1.1), the MX RRset is | |
790 | "secure", and the SMTP client MUST treat each MX hostname as a | |
791 | separate non-MX destination for opportunistic DANE TLS as described | |
792 | in Section 2.2.2. When, for a given MX hostname, no TLSA records are | |
793 | found, or only "insecure" TLSA records are found, DANE TLSA is not | |
794 | applicable with the SMTP server in question and delivery proceeds to | |
795 | that host as with pre-DANE opportunistic TLS. To avoid downgrade | |
796 | attacks, any errors during TLSA lookups MUST, as explained in | |
797 | Section 2.1.1, cause the SMTP server in question to be treated as | |
798 | unreachable. | |
799 | ||
800 | 2.2.2. Non-MX destinations | |
801 | ||
802 | This section describes the algorithm used to locate the TLSA records | |
803 | and associated TLSA base domain for an input domain not subject to MX | |
804 | resolution. Such domains include: | |
805 | ||
806 | o Each MX hostname used in a message delivery attempt for an | |
807 | original next-hop destination domain subject to MX resolution. | |
808 | Note, MTAs are not obligated to support CNAME expansion of MX | |
809 | hostnames. | |
810 | ||
811 | o Any administrator configured relay hostname, not subject to MX | |
812 | resolution. This frequently involves configuration set by the MTA | |
813 | administrator to handle some or all mail. | |
814 | ||
815 | o A next-hop destination domain subject to MX resolution that has no | |
816 | MX records. In this case the domain's name is implicitly also its | |
817 | sole SMTP server name. | |
818 | ||
819 | Note that DNS queries with type TLSA are mishandled by load balancing | |
820 | nameservers that serve the MX hostnames of some large email | |
821 | providers. The DNS zones served by these nameservers are not signed | |
822 | and contain no TLSA records, but queries for TLSA records fail, | |
823 | rather than returning the non-existence of the requested TLSA | |
824 | records. | |
825 | ||
826 | To avoid problems delivering mail to domains whose SMTP servers are | |
827 | served by the problem nameservers the SMTP client MUST perform any A | |
828 | and/or AAAA queries for the destination before attempting to locate | |
829 | the associated TLSA records. This lookup is needed in any case to | |
830 | determine whether the destination domain is reachable and the DNSSEC | |
831 | validation status of the chain of CNAME queries required to reach the | |
832 | ultimate address records. | |
833 | ||
834 | If no address records are found, the destination is unreachable. If | |
835 | address records are found, but the DNSSEC validation status of the | |
836 | first query response is "insecure" (see Section 2.1.3), the SMTP | |
837 | ||
838 | ||
839 | ||
840 | Dukhovni & Hardaker Expires November 26, 2014 [Page 15] | |
841 | \f | |
842 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
843 | ||
844 | ||
845 | client SHOULD NOT proceed to search for any associated TLSA records. | |
846 | With the problem domains, TLSA queries will lead to DNS lookup errors | |
847 | and cause messages to be consistently delayed and ultimately returned | |
848 | to the sender. We don't expect to find any "secure" TLSA records | |
849 | associated with a TLSA base domain that lies in an unsigned DNS zone. | |
850 | Therefore, skipping TLSA lookups in this case will also reduce | |
851 | latency with no detrimental impact on security. | |
852 | ||
853 | If the A and/or AAAA lookup of the "initial name" yields a CNAME, we | |
854 | replace it with the resulting name as if it were the initial name and | |
855 | perform a lookup again using the new name. This replacement is | |
856 | performed recursively. | |
857 | ||
858 | We consider the following cases for handling a DNS response for an A | |
859 | or AAAA DNS lookup: | |
860 | ||
861 | Not found: When the DNS queries for A and/or AAAA records yield | |
862 | neither a list of addresses nor a CNAME (or CNAME expansion is not | |
863 | supported) the destination is unreachable. | |
864 | ||
865 | Non-CNAME: The answer is not a CNAME alias. If the address RRset | |
866 | is "secure", TLSA lookups are performed as described in | |
867 | Section 2.2.3 with the initial name as the candidate TLSA base | |
868 | domain. If no "secure" TLSA records are found, DANE TLS is not | |
869 | applicable and mail delivery proceeds with pre-DANE opportunistic | |
870 | TLS (which, being best-effort, degrades to cleartext delivery when | |
871 | STARTTLS is not available or the TLS handshake fails). | |
872 | ||
873 | Insecure CNAME: The input domain is a CNAME alias, but the ultimate | |
874 | network address RRset is "insecure" (see Section 2.1.1). If the | |
875 | initial CNAME response is also "insecure", DANE TLS does not | |
876 | apply. Otherwise, this case is treated just like the non-CNAME | |
877 | case above, where a search is performed for a TLSA record with the | |
878 | original input domain as the candidate TLSA base domain. | |
879 | ||
880 | Secure CNAME: The input domain is a CNAME alias, and the ultimate | |
881 | network address RRset is "secure" (see Section 2.1.1). Two | |
882 | candidate TLSA base domains are tried: the fully CNAME-expanded | |
883 | initial name and, failing that, then the initial name itself. | |
884 | ||
885 | ||
886 | ||
887 | ||
888 | ||
889 | ||
890 | ||
891 | ||
892 | ||
893 | ||
894 | ||
895 | ||
896 | Dukhovni & Hardaker Expires November 26, 2014 [Page 16] | |
897 | \f | |
898 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
899 | ||
900 | ||
901 | In summary, if it is possible to securely obtain the full, CNAME- | |
902 | expanded, DNSSEC-validated address records for the input domain, then | |
903 | that name is the preferred TLSA base domain. Otherwise, the | |
904 | unexpanded input-MX domain is the candidate TLSA base domain. When | |
905 | no "secure" TLSA records are found at either the CNAME-expanded or | |
906 | unexpanded domain, then DANE TLS does not apply for mail delivery via | |
907 | the input domain in question. And, as always, errors, bogus or | |
908 | indeterminate results for any query in the process MUST result in | |
909 | delaying or abandoning delivery. | |
910 | ||
911 | 2.2.3. TLSA record lookup | |
912 | ||
913 | Each candidate TLSA base domain (the original or fully CNAME-expanded | |
914 | name of a non-MX destination or a particular MX hostname of an MX | |
915 | destination) is in turn prefixed with service labels of the form | |
916 | "_<port>._tcp". The resulting domain name is used to issue a DNSSEC | |
917 | query with the query type set to TLSA ([RFC6698] Section 7.1). | |
918 | ||
919 | For SMTP, the destination TCP port is typically 25, but this may be | |
920 | different with custom routes specified by the MTA administrator in | |
921 | which case the SMTP client MUST use the appropriate number in the | |
922 | "_<port>" prefix in place of "_25". If, for example, the candidate | |
923 | base domain is "mx.example.com", and the SMTP connection is to port | |
924 | 25, the TLSA RRset is obtained via a DNSSEC query of the form: | |
925 | ||
926 | _25._tcp.mx.example.com. IN TLSA ? | |
927 | ||
928 | The query response may be a CNAME, or the actual TLSA RRset. If the | |
929 | response is a CNAME, the SMTP client (through the use of its | |
930 | security-aware stub resolver) restarts the TLSA query at the target | |
931 | domain, following CNAMEs as appropriate and keeping track of whether | |
932 | the entire chain is "secure". If any "insecure" records are | |
933 | encountered, or the TLSA records don't exist, the next candidate TLSA | |
934 | base is tried instead. | |
935 | ||
936 | If the ultimate response is a "secure" TLSA RRset, then the candidate | |
937 | TLSA base domain will be the actual TLSA base domain and the TLSA | |
938 | RRset will constitute the TLSA records for the destination. If none | |
939 | of the candidate TLSA base domains yield "secure" TLSA records then | |
940 | delivery MAY proceed via pre-DANE opportunistic TLS. SMTP clients | |
941 | MAY elect to use "insecure" TLSA records to avoid STARTTLS downgrades | |
942 | or even to skip SMTP servers that fail authentication, but MUST NOT | |
943 | misrepresent authentication success as either a secure connection to | |
944 | the SMTP server or as a secure delivery to the intended next-hop | |
945 | domain. | |
946 | ||
947 | TLSA record publishers may leverage CNAMEs to reference a single | |
948 | authoritative TLSA RRset specifying a common Certification Authority | |
949 | ||
950 | ||
951 | ||
952 | Dukhovni & Hardaker Expires November 26, 2014 [Page 17] | |
953 | \f | |
954 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
955 | ||
956 | ||
957 | or a common end entity certificate to be used with multiple TLS | |
958 | services. Such CNAME expansion does not change the SMTP client's | |
959 | notion of the TLSA base domain; thus, when _25._tcp.mx.example.com is | |
960 | a CNAME, the base domain remains mx.example.com and this is still the | |
961 | reference identifier used together with the next-hop domain in peer | |
962 | certificate name checks. | |
963 | ||
964 | Note, shared end entity certificate associations expose the | |
965 | publishing domain to substitution attacks, where an MITM attacker can | |
966 | reroute traffic to a different server that shares the same end entity | |
967 | certificate. Such shared end entity records SHOULD be avoided unless | |
968 | the servers in question are functionally equivalent (an active | |
969 | attacker gains nothing by diverting client traffic from one such | |
970 | server to another). | |
971 | ||
972 | For example, given the DNSSEC validated records below: | |
973 | ||
974 | example.com. IN MX 0 mx1.example.com. | |
975 | example.com. IN MX 0 mx2.example.com. | |
976 | _25._tcp.mx1.example.com. IN CNAME tlsa211._dane.example.com. | |
977 | _25._tcp.mx2.example.com. IN CNAME tlsa211._dane.example.com. | |
978 | tlsa211._dane.example.com. IN TLSA 2 1 1 e3b0c44298fc1c149a... | |
979 | ||
980 | The SMTP servers mx1.example.com and mx2.example.com will be expected | |
981 | to have certificates issued under a common trust anchor, but each MX | |
982 | hostname's TLSA base domain remains unchanged despite the above CNAME | |
983 | records. Correspondingly, each SMTP server will be associated with a | |
984 | pair of reference identifiers consisting of its hostname plus the | |
985 | next-hop domain "example.com". | |
986 | ||
987 | If, during TLSA resolution (including possible CNAME indirection), at | |
988 | least one "secure" TLSA record is found (even if not usable because | |
989 | it is unsupported by the implementation or support is | |
990 | administratively disabled), then the corresponding host has signaled | |
991 | its commitment to implement TLS. The SMTP client MUST NOT deliver | |
992 | mail via the corresponding host unless a TLS session is negotiated | |
993 | via STARTTLS. This is required to avoid MITM STARTTLS downgrade | |
994 | attacks. | |
995 | ||
996 | As noted previously (in Section Section 2.2.2), when no "secure" TLSA | |
997 | records are found at the fully CNAME-expanded name, the original | |
998 | unexpanded name MUST be tried instead. This supports customers of | |
999 | hosting providers where the provider's zone cannot be validated with | |
1000 | DNSSEC, but the customer has shared appropriate key material with the | |
1001 | hosting provider to enable TLS via SNI. Intermediate names that | |
1002 | arise during CNAME expansion that are neither the original, nor the | |
1003 | final name, are never candidate TLSA base domains, even if "secure". | |
1004 | ||
1005 | ||
1006 | ||
1007 | ||
1008 | Dukhovni & Hardaker Expires November 26, 2014 [Page 18] | |
1009 | \f | |
1010 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1011 | ||
1012 | ||
1013 | 3. DANE authentication | |
1014 | ||
1015 | This section describes which TLSA records are applicable to SMTP | |
1016 | opportunistic DANE TLS and how to apply such records to authenticate | |
1017 | the SMTP server. With opportunistic DANE TLS, both the TLS support | |
1018 | implied by the presence of DANE TLSA records and the verification | |
1019 | parameters necessary to authenticate the TLS peer are obtained | |
1020 | together. In contrast to protocols where channel security policy is | |
1021 | set exclusively by the client, authentication via this protocol is | |
1022 | expected to be less prone to connection failure caused by | |
1023 | incompatible configuration of the client and server. | |
1024 | ||
1025 | 3.1. TLSA certificate usages | |
1026 | ||
1027 | The DANE TLSA specification [RFC6698] defines multiple TLSA RR types | |
1028 | via combinations of 3 numeric parameters. The numeric values of | |
1029 | these parameters were later given symbolic names in | |
1030 | [I-D.ietf-dane-registry-acronyms]. The rest of the TLSA record is | |
1031 | the "certificate association data field", which specifies the full or | |
1032 | digest value of a certificate or public key. The parameters are: | |
1033 | ||
1034 | The TLSA Certificate Usage field: Section 2.1.1 of [RFC6698] | |
1035 | specifies 4 values: PKIX-TA(0), PKIX-EE(1), DANE-TA(2), and DANE- | |
1036 | EE(3). There is an additional private-use value: PrivCert(255). | |
1037 | All other values are reserved for use by future specifications. | |
1038 | ||
1039 | The selector field: Section 2.1.2 of [RFC6698] specifies 2 values: | |
1040 | Cert(0), SPKI(1). There is an additional private-use value: | |
1041 | PrivSel(255). All other values are reserved for use by future | |
1042 | specifications. | |
1043 | ||
1044 | The matching type field: Section 2.1.3 of [RFC6698] specifies 3 | |
1045 | values: Full(0), SHA2-256(1), SHA2-512(2). There is an additional | |
1046 | private-use value: PrivMatch(255). All other values are reserved | |
1047 | for use by future specifications. | |
1048 | ||
1049 | We may think of TLSA Certificate Usage values 0 through 3 as a | |
1050 | combination of two one-bit flags. The low bit chooses between trust | |
1051 | anchor (TA) and end entity (EE) certificates. The high bit chooses | |
1052 | between public PKI issued and domain-issued certificates. | |
1053 | ||
1054 | The selector field specifies whether the TLSA RR matches the whole | |
1055 | certificate: Cert(0), or just its subjectPublicKeyInfo: SPKI(1). The | |
1056 | subjectPublicKeyInfo is an ASN.1 DER encoding of the certificate's | |
1057 | algorithm id, any parameters and the public key data. | |
1058 | ||
1059 | The matching type field specifies how the TLSA RR Certificate | |
1060 | Association Data field is to be compared with the certificate or | |
1061 | ||
1062 | ||
1063 | ||
1064 | Dukhovni & Hardaker Expires November 26, 2014 [Page 19] | |
1065 | \f | |
1066 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1067 | ||
1068 | ||
1069 | public key. A value of Full(0) means an exact match: the full DER | |
1070 | encoding of the certificate or public key is given in the TLSA RR. A | |
1071 | value of SHA2-256(1) means that the association data matches the | |
1072 | SHA2-256 digest of the certificate or public key, and likewise | |
1073 | SHA2-512(2) means a SHA2-512 digest is used. | |
1074 | ||
1075 | Since opportunistic DANE TLS will be used by non-interactive MTAs, | |
1076 | with no user to "press OK" when authentication fails, reliability of | |
1077 | peer authentication is paramount. Server operators are advised to | |
1078 | publish TLSA records that are least likely to fail authentication due | |
1079 | to interoperability or operational problems. Because DANE TLS relies | |
1080 | on coordinated changes to DNS and SMTP server settings, the best | |
1081 | choice of records to publish will depend on site-specific practices. | |
1082 | ||
1083 | The certificate usage element of a TLSA record plays a critical role | |
1084 | in determining how the corresponding certificate association data | |
1085 | field is used to authenticate server's certificate chain. The next | |
1086 | two subsections explain the process for certificate usages DANE-EE(3) | |
1087 | and DANE-TA(2). The third subsection briefly explains why | |
1088 | certificate usages PKIX-TA(0) and PKIX-EE(1) are not applicable with | |
1089 | opportunistic DANE TLS. | |
1090 | ||
1091 | In summary, we recommend the use of either "DANE-EE(3) SPKI(1) | |
1092 | SHA2-256(1)" or "DANE-TA(2) Cert(0) SHA2-256(1)" TLSA records | |
1093 | depending on site needs. Other combinations of TLSA parameters are | |
1094 | either explicitly unsupported, or offer little to recommend them over | |
1095 | these two. | |
1096 | ||
1097 | The mandatory to support digest algorithm in [RFC6698] is | |
1098 | SHA2-256(1). When the server's TLSA RRset includes records with a | |
1099 | matching type indicating a digest record (i.e., a value other than | |
1100 | Full(0)), a TLSA record with a SHA2-256(1) matching type SHOULD be | |
1101 | provided along with any other digest published, since some SMTP | |
1102 | clients may support only SHA2-256(1). If at some point the SHA2-256 | |
1103 | digest algorithm is tarnished by new cryptanalytic attacks, | |
1104 | publishers will need to include an appropriate stronger digest in | |
1105 | their TLSA records, initially along with, and ultimately in place of, | |
1106 | SHA2-256. | |
1107 | ||
1108 | 3.1.1. Certificate usage DANE-EE(3) | |
1109 | ||
1110 | Authentication via certificate usage DANE-EE(3) TLSA records involves | |
1111 | simply checking that the server's leaf certificate matches the TLSA | |
1112 | record. In particular the binding of the server public key to its | |
1113 | name is based entirely on the TLSA record association. The server | |
1114 | MUST be considered authenticated even if none of the names in the | |
1115 | certificate match the client's reference identity for the server. | |
1116 | ||
1117 | ||
1118 | ||
1119 | ||
1120 | Dukhovni & Hardaker Expires November 26, 2014 [Page 20] | |
1121 | \f | |
1122 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1123 | ||
1124 | ||
1125 | Similarly, the expiration date of the server certificate MUST be | |
1126 | ignored, the validity period of the TLSA record key binding is | |
1127 | determined by the validity interval of the TLSA record DNSSEC | |
1128 | signature. | |
1129 | ||
1130 | With DANE-EE(3) servers need not employ SNI (may ignore the client's | |
1131 | SNI message) even when the server is known under independent names | |
1132 | that would otherwise require separate certificates. It is instead | |
1133 | sufficient for the TLSA RRsets for all the domains in question to | |
1134 | match the server's default certificate. Of course with SMTP servers | |
1135 | it is simpler still to publish the same MX hostname for all the | |
1136 | hosted domains. | |
1137 | ||
1138 | For domains where it is practical to make coordinated changes in DNS | |
1139 | TLSA records during SMTP server key rotation, it is often best to | |
1140 | publish end-entity DANE-EE(3) certificate associations. DANE-EE(3) | |
1141 | certificates don't suddenly stop working when leaf or intermediate | |
1142 | certificates expire, and don't fail when the server operator neglects | |
1143 | to configure all the required issuer certificates in the server | |
1144 | certificate chain. | |
1145 | ||
1146 | TLSA records published for SMTP servers SHOULD, in most cases, be | |
1147 | "DANE-EE(3) SPKI(1) SHA2-256(1)" records. Since all DANE | |
1148 | implementations are required to support SHA2-256, this record type | |
1149 | works for all clients and need not change across certificate renewals | |
1150 | with the same key. | |
1151 | ||
1152 | 3.1.2. Certificate usage DANE-TA(2) | |
1153 | ||
1154 | Some domains may prefer to avoid the operational complexity of | |
1155 | publishing unique TLSA RRs for each TLS service. If the domain | |
1156 | employs a common issuing Certification Authority to create | |
1157 | certificates for multiple TLS services, it may be simpler to publish | |
1158 | the issuing authority as a trust anchor (TA) for the certificate | |
1159 | chains of all relevant services. The TLSA query domain (TLSA base | |
1160 | domain with port and protocol prefix labels) for each service issued | |
1161 | by the same TA may then be set to a CNAME alias that points to a | |
1162 | common TLSA RRset that matches the TA. For example: | |
1163 | ||
1164 | example.com. IN MX 0 mx1.example.com. | |
1165 | example.com. IN MX 0 mx2.example.com. | |
1166 | _25._tcp.mx1.example.com. IN CNAME tlsa211._dane.example.com. | |
1167 | _25._tcp.mx2.example.com. IN CNAME tlsa211._dane.example.com. | |
1168 | tlsa211._dane.example.com. IN TLSA 2 1 1 e3b0c44298fc1c14.... | |
1169 | ||
1170 | ||
1171 | ||
1172 | ||
1173 | ||
1174 | ||
1175 | ||
1176 | Dukhovni & Hardaker Expires November 26, 2014 [Page 21] | |
1177 | \f | |
1178 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1179 | ||
1180 | ||
1181 | With usage DANE-TA(2) the server certificates will need to have names | |
1182 | that match one of the client's reference identifiers (see [RFC6125]). | |
1183 | The server MAY employ SNI to select the appropriate certificate to | |
1184 | present to the client. | |
1185 | ||
1186 | SMTP servers that rely on certificate usage DANE-TA(2) TLSA records | |
1187 | for TLS authentication MUST include the TA certificate as part of the | |
1188 | certificate chain presented in the TLS handshake server certificate | |
1189 | message even when it is a self-signed root certificate. At this | |
1190 | time, many SMTP servers are not configured with a comprehensive list | |
1191 | of trust anchors, nor are they expected to at any point in the | |
1192 | future. Some MTAs will ignore all locally trusted certificates when | |
1193 | processing usage DANE-TA(2) TLSA records. Thus even when the TA | |
1194 | happens to be a public Certification Authority known to the SMTP | |
1195 | client, authentication is likely to fail unless the TA certificate is | |
1196 | included in the TLS server certificate message. | |
1197 | ||
1198 | TLSA records with selector Full(0) are discouraged. While these | |
1199 | potentially obviate the need to transmit the TA certificate in the | |
1200 | TLS server certificate message, client implementations may not be | |
1201 | able to augment the server certificate chain with the data obtained | |
1202 | from DNS, especially when the TLSA record supplies a bare key | |
1203 | (selector SPKI(1)). Since the server will need to transmit the TA | |
1204 | certificate in any case, server operators SHOULD publish TLSA records | |
1205 | with a selector other than Full(0) and avoid potential | |
1206 | interoperability issues with large TLSA records containing full | |
1207 | certificates or keys. | |
1208 | ||
1209 | TLSA Publishers employing DANE-TA(2) records SHOULD publish records | |
1210 | with a selector of Cert(0). Such TLSA records are associated with | |
1211 | the whole trust anchor certificate, not just with the trust anchor | |
1212 | public key. In particular, the SMTP client SHOULD then apply any | |
1213 | relevant constraints from the trust anchor certificate, such as, for | |
1214 | example, path length constraints. | |
1215 | ||
1216 | While a selector of SPKI(1) may also be employed, the resulting TLSA | |
1217 | record will not specify the full trust anchor certificate content, | |
1218 | and elements of the trust anchor certificate other than the public | |
1219 | key become mutable. This may, for example, allow a subsidiary CA to | |
1220 | issue a chain that violates the trust anchor's path length or name | |
1221 | constraints. | |
1222 | ||
1223 | 3.1.3. Certificate usages PKIX-TA(0) and PKIX-EE(1) | |
1224 | ||
1225 | As noted in the introduction, SMTP clients cannot, without relying on | |
1226 | DNSSEC for secure MX records and DANE for STARTTLS support signaling, | |
1227 | perform server identity verification or prevent STARTTLS downgrade | |
1228 | attacks. The use of PKIX CAs offers no added security since an | |
1229 | ||
1230 | ||
1231 | ||
1232 | Dukhovni & Hardaker Expires November 26, 2014 [Page 22] | |
1233 | \f | |
1234 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1235 | ||
1236 | ||
1237 | attacker capable of compromising DNSSEC is free to replace any PKIX- | |
1238 | TA(0) or PKIX-EE(1) TLSA records with records bearing any convenient | |
1239 | non-PKIX certificate usage. | |
1240 | ||
1241 | SMTP servers SHOULD NOT publish TLSA RRs with certificate usage PKIX- | |
1242 | TA(0) or PKIX-EE(1). SMTP clients cannot be expected to be | |
1243 | configured with a suitably complete set of trusted public CAs. | |
1244 | Lacking a complete set of public CAs, clients would not be able to | |
1245 | verify the certificates of SMTP servers whose issuing root CAs are | |
1246 | not trusted by the client. | |
1247 | ||
1248 | Opportunistic DANE TLS needs to interoperate without bilateral | |
1249 | coordination of security settings between client and server systems. | |
1250 | Therefore, parameter choices that are fragile in the absence of | |
1251 | bilateral coordination are unsupported. Nothing is lost since the | |
1252 | PKIX certificate usages cannot aid SMTP TLS security, they can only | |
1253 | impede SMTP TLS interoperability. | |
1254 | ||
1255 | SMTP client treatment of TLSA RRs with certificate usages PKIX-TA(0) | |
1256 | or PKIX-EE(1) is undefined. SMTP clients should generally treat such | |
1257 | TLSA records as unusable. | |
1258 | ||
1259 | 3.2. Certificate matching | |
1260 | ||
1261 | When at least one usable "secure" TLSA record is found, the SMTP | |
1262 | client MUST use TLSA records to authenticate the SMTP server. | |
1263 | Messages MUST NOT be delivered via the SMTP server if authentication | |
1264 | fails, otherwise the SMTP client is vulnerable to MITM attacks. | |
1265 | ||
1266 | 3.2.1. DANE-EE(3) name checks | |
1267 | ||
1268 | The SMTP client MUST NOT perform certificate name checks with | |
1269 | certificate usage DANE-EE(3), see Section 3.1.1 above. | |
1270 | ||
1271 | 3.2.2. DANE-TA(2) name checks | |
1272 | ||
1273 | To match a server via a TLSA record with certificate usage DANE- | |
1274 | TA(2), the client MUST perform name checks to ensure that it has | |
1275 | reached the correct server. In all DANE-TA(2) cases the SMTP client | |
1276 | MUST include the TLSA base domain as one of the valid reference | |
1277 | identifiers for matching the server certificate. | |
1278 | ||
1279 | TLSA records for MX hostnames: If the TLSA base domain was obtained | |
1280 | indirectly via a "secure" MX lookup (including any CNAME-expanded | |
1281 | name of an MX hostname), then the original next-hop domain used in | |
1282 | the MX lookup MUST be included as as a second reference | |
1283 | identifier. The CNAME-expanded original next-hop domain MUST be | |
1284 | included as a third reference identifier if different from the | |
1285 | ||
1286 | ||
1287 | ||
1288 | Dukhovni & Hardaker Expires November 26, 2014 [Page 23] | |
1289 | \f | |
1290 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1291 | ||
1292 | ||
1293 | original next-hop domain. When the client MTA is employing DANE | |
1294 | TLS security despite "insecure" MX redirection the MX hostname is | |
1295 | the only reference identifier. | |
1296 | ||
1297 | TLSA records for Non-MX hostnames: If MX records were not used | |
1298 | (e.g., if none exist) and the TLSA base domain is the CNAME- | |
1299 | expanded original next-hop domain, then the original next-hop | |
1300 | domain MUST be included as a second reference identifier. | |
1301 | ||
1302 | Accepting certificates with the original next-hop domain in addition | |
1303 | to the MX hostname allows a domain with multiple MX hostnames to | |
1304 | field a single certificate bearing a single domain name (i.e., the | |
1305 | email domain) across all the SMTP servers. This also aids | |
1306 | interoperability with pre-DANE SMTP clients that are configured to | |
1307 | look for the email domain name in server certificates. For example, | |
1308 | with "secure" DNS records as below: | |
1309 | ||
1310 | exchange.example.org. IN CNAME mail.example.org. | |
1311 | mail.example.org. IN CNAME example.com. | |
1312 | example.com. IN MX 10 mx10.example.com. | |
1313 | example.com. IN MX 15 mx15.example.com. | |
1314 | example.com. IN MX 20 mx20.example.com. | |
1315 | ; | |
1316 | mx10.example.com. IN A 192.0.2.10 | |
1317 | _25._tcp.mx10.example.com. IN TLSA 2 0 1 ... | |
1318 | ; | |
1319 | mx15.example.com. IN CNAME mxbackup.example.com. | |
1320 | mxbackup.example.com. IN A 192.0.2.15 | |
1321 | ; _25._tcp.mxbackup.example.com. IN TLSA ? (NXDOMAIN) | |
1322 | _25._tcp.mx15.example.com. IN TLSA 2 0 1 ... | |
1323 | ; | |
1324 | mx20.example.com. IN CNAME mxbackup.example.net. | |
1325 | mxbackup.example.net. IN A 198.51.100.20 | |
1326 | _25._tcp.mxbackup.example.net. IN TLSA 2 0 1 ... | |
1327 | ||
1328 | Certificate name checks for delivery of mail to exchange.example.org | |
1329 | via any of the associated SMTP servers MUST accept at least the names | |
1330 | "exchange.example.org" and "example.com", which are respectively the | |
1331 | original and fully expanded next-hop domain. When the SMTP server is | |
1332 | mx10.example.com, name checks MUST accept the TLSA base domain | |
1333 | "mx10.example.com". If, despite the fact that MX hostnames are | |
1334 | required to not be aliases, the MTA supports delivery via | |
1335 | "mx15.example.com" or "mx20.example.com" then name checks MUST accept | |
1336 | the respective TLSA base domains "mx15.example.com" and | |
1337 | "mxbackup.example.net". | |
1338 | ||
1339 | 3.2.3. Reference identifier matching | |
1340 | ||
1341 | ||
1342 | ||
1343 | ||
1344 | Dukhovni & Hardaker Expires November 26, 2014 [Page 24] | |
1345 | \f | |
1346 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1347 | ||
1348 | ||
1349 | When name checks are applicable (certificate usage DANE-TA(2)), if | |
1350 | the server certificate contains a Subject Alternative Name extension | |
1351 | ([RFC5280]), with at least one DNS-ID ([RFC6125]) then only the DNS- | |
1352 | IDs are matched against the client's reference identifiers. The CN- | |
1353 | ID ([RFC6125]) is only considered when no DNS-IDs are present. The | |
1354 | server certificate is considered matched when one of its presented | |
1355 | identifiers ([RFC5280]) matches any of the client's reference | |
1356 | identifiers. | |
1357 | ||
1358 | Wildcards are valid in either DNS-IDs or the CN-ID when applicable. | |
1359 | The wildcard character must be entire first label of the DNS-ID or | |
1360 | CN-ID. Thus, "*.example.com" is valid, while "smtp*.example.com" and | |
1361 | "*smtp.example.com" are not. SMTP clients MUST support wildcards | |
1362 | that match the first label of the reference identifier, with the | |
1363 | remaining labels matching verbatim. For example, the DNS-ID | |
1364 | "*.example.com" matches the reference identifier "mx1.example.com". | |
1365 | SMTP clients MAY, subject to local policy allow wildcards to match | |
1366 | multiple reference identifier labels, but servers cannot expect broad | |
1367 | support for such a policy. Therefore any wildcards in server | |
1368 | certificates SHOULD match exactly one label in either the TLSA base | |
1369 | domain or the next-hop domain. | |
1370 | ||
1371 | 4. Server key management | |
1372 | ||
1373 | Two TLSA records MUST be published before employing a new EE or TA | |
1374 | public key or certificate, one matching the currently deployed key | |
1375 | and the other matching the new key scheduled to replace it. Once | |
1376 | sufficient time has elapsed for all DNS caches to expire the previous | |
1377 | TLSA RRset and related signature RRsets, servers may be configured to | |
1378 | use the new EE private key and associated public key certificate or | |
1379 | may employ certificates signed by the new trust anchor. | |
1380 | ||
1381 | Once the new public key or certificate is in use, the TLSA RR that | |
1382 | matches the retired key can be removed from DNS, leaving only RRs | |
1383 | that match keys or certificates in active use. | |
1384 | ||
1385 | As described in Section 3.1.2, when server certificates are validated | |
1386 | via a DANE-TA(2) trust anchor, and CNAME records are employed to | |
1387 | store the TA association data at a single location, the | |
1388 | responsibility of updating the TLSA RRset shifts to the operator of | |
1389 | the trust anchor. Before a new trust anchor is used to sign any new | |
1390 | server certificates, its certificate (digest) is added to the | |
1391 | relevant TLSA RRset. After enough time elapses for the original TLSA | |
1392 | RRset to age out of DNS caches, the new trust anchor can start | |
1393 | issuing new server certificates. Once all certificates issued under | |
1394 | the previous trust anchor have expired, its associated RRs can be | |
1395 | removed from the TLSA RRset. | |
1396 | ||
1397 | ||
1398 | ||
1399 | ||
1400 | Dukhovni & Hardaker Expires November 26, 2014 [Page 25] | |
1401 | \f | |
1402 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1403 | ||
1404 | ||
1405 | In the DANE-TA(2) key management model server operators do not | |
1406 | generally need to update DNS TLSA records after initially creating a | |
1407 | CNAME record that references the centrally operated DANE-TA(2) RRset. | |
1408 | If a particular server's key is compromised, its TLSA CNAME SHOULD be | |
1409 | replaced with a DANE-EE(3) association until the certificate for the | |
1410 | compromised key expires, at which point it can return to using CNAME | |
1411 | record. If the central trust anchor is compromised, all servers need | |
1412 | to be issued new keys by a new TA, and a shared DANE-TA(2) TLSA RRset | |
1413 | needs to be published containing just the new TA. SMTP servers | |
1414 | cannot expect broad SMTP client CRL or OCSP support. | |
1415 | ||
1416 | 5. Digest algorithm agility | |
1417 | ||
1418 | While [RFC6698] specifies multiple digest algorithms, it does not | |
1419 | specify a protocol by which the SMTP client and TLSA record publisher | |
1420 | can agree on the strongest shared algorithm. Such a protocol would | |
1421 | allow the client and server to avoid exposure to any deprecated | |
1422 | weaker algorithms that are published for compatibility with less | |
1423 | capable clients, but should be ignored when possible. We specify | |
1424 | such a protocol below. | |
1425 | ||
1426 | Suppose that a DANE TLS client authenticating a TLS server considers | |
1427 | digest algorithm "BetterAlg" stronger than digest algorithm | |
1428 | "WorseAlg". Suppose further that a server's TLSA RRset contains some | |
1429 | records with "BetterAlg" as the digest algorithm. Finally, suppose | |
1430 | that for every raw public key or certificate object that is included | |
1431 | in the server's TLSA RRset in digest form, whenever that object | |
1432 | appears with algorithm "WorseAlg" with some usage and selector it | |
1433 | also appears with algorithm "BetterAlg" with the same usage and | |
1434 | selector. In that case our client can safely ignore TLSA records | |
1435 | with the weaker algorithm "WorseAlg", because it suffices to check | |
1436 | the records with the stronger algorithm "BetterAlg". | |
1437 | ||
1438 | Server operators MUST ensure that for any given usage and selector, | |
1439 | each object (certificate or public key), for which a digest | |
1440 | association exists in the TLSA RRset, is published with the SAME SET | |
1441 | of digest algorithms as all other objects that published with that | |
1442 | usage and selector. In other words, for each usage and selector, the | |
1443 | records with non-zero matching types will correspond to on a cross- | |
1444 | product of a set of underlying objects and a fixed set of digest | |
1445 | algorithms that apply uniformly to all the objects. | |
1446 | ||
1447 | To achieve digest algorithm agility, all published TLSA RRsets for | |
1448 | use with opportunistic DANE TLS for SMTP MUST conform to the above | |
1449 | requirements. Then, for each combination of usage and selector, SMTP | |
1450 | clients can simply ignore all digest records except those that employ | |
1451 | the strongest digest algorithm. The ordering of digest algorithms by | |
1452 | strength is not specified in advance, it is entirely up to the SMTP | |
1453 | ||
1454 | ||
1455 | ||
1456 | Dukhovni & Hardaker Expires November 26, 2014 [Page 26] | |
1457 | \f | |
1458 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1459 | ||
1460 | ||
1461 | client. SMTP client implementations SHOULD make the digest algorithm | |
1462 | preference order configurable. Only the future will tell which | |
1463 | algorithms might be weakened by new attacks and when. | |
1464 | ||
1465 | Note, TLSA records with a matching type of Full(0), that publish the | |
1466 | full value of a certificate or public key object, play no role in | |
1467 | digest algorithm agility. They neither trump the processing of | |
1468 | records that employ digests, nor are they ignored in the presence of | |
1469 | any records with a digest (i.e. non-zero) matching type. | |
1470 | ||
1471 | SMTP clients SHOULD use digest algorithm agility when processing the | |
1472 | DANE TLSA records of an SMTP server. Algorithm agility is to be | |
1473 | applied after first discarding any unusable or malformed records | |
1474 | (unsupported digest algorithm, or incorrect digest length). Thus, | |
1475 | for each usage and selector, the client SHOULD process only any | |
1476 | usable records with a matching type of Full(0) and the usable records | |
1477 | whose digest algorithm is believed to be the strongest among usable | |
1478 | records with the given usage and selector. | |
1479 | ||
1480 | The main impact of this requirement is on key rotation, when the TLSA | |
1481 | RRset is pre-populated with digests of new certificates or public | |
1482 | keys, before these replace or augment their predecessors. Were the | |
1483 | newly introduced RRs to include previously unused digest algorithms, | |
1484 | clients that employ this protocol could potentially ignore all the | |
1485 | digests corresponding to the current keys or certificates, causing | |
1486 | connectivity issues until the new keys or certificates are deployed. | |
1487 | Similarly, publishing new records with fewer digests could cause | |
1488 | problems for clients using cached TLSA RRsets that list both the old | |
1489 | and new objects once the new keys are deployed. | |
1490 | ||
1491 | To avoid problems, server operators SHOULD apply the following | |
1492 | strategy: | |
1493 | ||
1494 | o When changing the set of objects published via the TLSA RRset | |
1495 | (e.g. during key rotation), DO NOT change the set of digest | |
1496 | algorithms used; change just the list of objects. | |
1497 | ||
1498 | o When changing the set of digest algorithms, change only the set of | |
1499 | algorithms, and generate a new RRset in which all the current | |
1500 | objects are re-published with the new set of digest algorithms. | |
1501 | ||
1502 | After either of these two changes are made, the new TLSA RRset should | |
1503 | be left in place long enough that the older TLSA RRset can be flushed | |
1504 | from caches before making another change. | |
1505 | ||
1506 | 6. Mandatory TLS Security | |
1507 | ||
1508 | ||
1509 | ||
1510 | ||
1511 | ||
1512 | Dukhovni & Hardaker Expires November 26, 2014 [Page 27] | |
1513 | \f | |
1514 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1515 | ||
1516 | ||
1517 | An MTA implementing this protocol may require a stronger security | |
1518 | assurance when sending email to selected destinations. The sending | |
1519 | organization may need to send sensitive email and/or may have | |
1520 | regulatory obligations to protect its content. This protocol is not | |
1521 | in conflict with such a requirement, and in fact can often simplify | |
1522 | authenticated delivery to such destinations. | |
1523 | ||
1524 | Specifically, with domains that publish DANE TLSA records for their | |
1525 | MX hostnames, a sending MTA can be configured to use the receiving | |
1526 | domains's DANE TLSA records to authenticate the corresponding SMTP | |
1527 | server. Authentication via DANE TLSA records is easier to manage, as | |
1528 | changes in the receiver's expected certificate properties are made on | |
1529 | the receiver end and don't require manually communicated | |
1530 | configuration changes. With mandatory DANE TLS, when no usable TLSA | |
1531 | records are found, message delivery is delayed. Thus, mail is only | |
1532 | sent when an authenticated TLS channel is established to the remote | |
1533 | SMTP server. | |
1534 | ||
1535 | Administrators of mail servers that employ mandatory DANE TLS, need | |
1536 | to carefully monitor their mail logs and queues. If a partner domain | |
1537 | unwittingly misconfigures their TLSA records, disables DNSSEC, or | |
1538 | misconfigures SMTP server certificate chains, mail will be delayed | |
1539 | and may bounce if the issue is not resolved in a timely manner. | |
1540 | ||
1541 | 7. Note on DANE for Message User Agents | |
1542 | ||
1543 | We note that the SMTP protocol is also used between Message User | |
1544 | Agents (MUAs) and Message Submission Agents (MSAs) [RFC6409]. In | |
1545 | [RFC6186] a protocol is specified that enables an MUA to dynamically | |
1546 | locate the MSA based on the user's email address. SMTP connection | |
1547 | security considerations for MUAs implementing [RFC6186] are largely | |
1548 | analogous to connection security requirements for MTAs, and this | |
1549 | specification could be applied largely verbatim with DNS MX records | |
1550 | replaced by corresponding DNS Service (SRV) records | |
1551 | [I-D.ietf-dane-srv]. | |
1552 | ||
1553 | However, until MUAs begin to adopt the dynamic configuration | |
1554 | mechanisms of [RFC6186] they are adequately served by more | |
1555 | traditional static TLS security policies. Specification of DANE TLS | |
1556 | for Message User Agent (MUA) to Message Submission Agent (MSA) SMTP | |
1557 | is left to future documents that focus specifically on SMTP security | |
1558 | between MUAs and MSAs. | |
1559 | ||
1560 | ||
1561 | ||
1562 | ||
1563 | ||
1564 | ||
1565 | ||
1566 | ||
1567 | ||
1568 | Dukhovni & Hardaker Expires November 26, 2014 [Page 28] | |
1569 | \f | |
1570 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1571 | ||
1572 | ||
1573 | 8. Interoperability considerations | |
1574 | ||
1575 | 8.1. SNI support | |
1576 | ||
1577 | To ensure that the server sends the right certificate chain, the SMTP | |
1578 | client MUST send the TLS SNI extension containing the TLSA base | |
1579 | domain. This precludes the use of the backward compatible SSL 2.0 | |
1580 | compatible SSL HELLO by the SMTP client. The minimum SSL/TLS client | |
1581 | HELLO version for SMTP clients performing DANE authentication is SSL | |
1582 | 3.0, but a client that offers SSL 3.0 MUST also offer at least TLS | |
1583 | 1.0 and MUST include the SNI extension. Servers that don't make use | |
1584 | of SNI MAY negotiate SSL 3.0 if offered by the client. | |
1585 | ||
1586 | Each SMTP server MUST present a certificate chain (see [RFC5246] | |
1587 | Section 7.4.2) that matches at least one of the TLSA records. The | |
1588 | server MAY rely on SNI to determine which certificate chain to | |
1589 | present to the client. Clients that don't send SNI information may | |
1590 | not see the expected certificate chain. | |
1591 | ||
1592 | If the server's TLSA records match the server's default certificate | |
1593 | chain, the server need not support SNI. In either case, the server | |
1594 | need not include the SNI extension in its TLS HELLO as simply | |
1595 | returning a matching certificate chain is sufficient. Servers MUST | |
1596 | NOT enforce the use of SNI by clients, as the client may be using | |
1597 | unauthenticated opportunistic TLS and may not expect any particular | |
1598 | certificate from the server. If the client sends no SNI extension, | |
1599 | or sends an SNI extension for an unsupported domain, the server MUST | |
1600 | simply send some fallback certificate chain of its choice. The | |
1601 | reason for not enforcing strict matching of the requested SNI | |
1602 | hostname is that DANE TLS clients are typically willing to accept | |
1603 | multiple server names, but can only send one name in the SNI | |
1604 | extension. The server's fallback certificate may match a different | |
1605 | name acceptable to the client, e.g., the original next-hop domain. | |
1606 | ||
1607 | 8.2. Anonymous TLS cipher suites | |
1608 | ||
1609 | Since many SMTP servers either do not support or do not enable any | |
1610 | anonymous TLS cipher suites, SMTP client TLS HELLO messages SHOULD | |
1611 | offer to negotiate a typical set of non-anonymous cipher suites | |
1612 | required for interoperability with such servers. An SMTP client | |
1613 | employing pre-DANE opportunistic TLS MAY in addition include one or | |
1614 | more anonymous TLS cipher suites in its TLS HELLO. SMTP servers, | |
1615 | that need to interoperate with opportunistic TLS clients SHOULD be | |
1616 | prepared to interoperate with such clients by either always selecting | |
1617 | a mutually supported non-anonymous cipher suite or by correctly | |
1618 | handling client connections that negotiate anonymous cipher suites. | |
1619 | ||
1620 | ||
1621 | ||
1622 | ||
1623 | ||
1624 | Dukhovni & Hardaker Expires November 26, 2014 [Page 29] | |
1625 | \f | |
1626 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1627 | ||
1628 | ||
1629 | Note that while SMTP server operators are under no obligation to | |
1630 | enable anonymous cipher suites, no security is gained by sending | |
1631 | certificates to clients that will ignore them. Indeed support for | |
1632 | anonymous cipher suites in the server makes audit trails more | |
1633 | informative. Log entries that record connections that employed an | |
1634 | anonymous cipher suite record the fact that the clients did not care | |
1635 | to authenticate the server. | |
1636 | ||
1637 | 9. Operational Considerations | |
1638 | ||
1639 | 9.1. Client Operational Considerations | |
1640 | ||
1641 | An operational error on the sending or receiving side that cannot be | |
1642 | corrected in a timely manner may, at times, lead to consistent | |
1643 | failure to deliver time-sensitive email. The sending MTA | |
1644 | administrator may have to choose between letting email queue until | |
1645 | the error is resolved and disabling opportunistic or mandatory DANE | |
1646 | TLS for one or more destinations. The choice to disable DANE TLS | |
1647 | security should not be made lightly. Every reasonable effort should | |
1648 | be made to determine that problems with mail delivery are the result | |
1649 | of an operational error, and not an attack. A fallback strategy may | |
1650 | be to configure explicit out-of-band TLS security settings if | |
1651 | supported by the sending MTA. | |
1652 | ||
1653 | SMTP clients may deploy opportunistic DANE TLS incrementally by | |
1654 | enabling it only for selected sites, or may occasionally need to | |
1655 | disable opportunistic DANE TLS for peers that fail to interoperate | |
1656 | due to misconfiguration or software defects on either end. Some | |
1657 | implementations MAY support DANE TLS in an "audit only" mode in which | |
1658 | failure to achieve the requisite security level is logged as a | |
1659 | warning and delivery proceeds at a reduced security level. Unless | |
1660 | local policy specifies "audit only" or that opportunistic DANE TLS is | |
1661 | not to be used for a particular destination, an SMTP client MUST NOT | |
1662 | deliver mail via a server whose certificate chain fails to match at | |
1663 | least one TLSA record when usable TLSA records are found for that | |
1664 | server. | |
1665 | ||
1666 | 9.2. Publisher Operational Considerations | |
1667 | ||
1668 | SMTP servers that publish certificate usage DANE-TA(2) associations | |
1669 | MUST include the TA certificate in their TLS server certificate | |
1670 | chain, even when that TA certificate is a self-signed root | |
1671 | certificate. | |
1672 | ||
1673 | TLSA Publishers must follow the digest agility guidelines in | |
1674 | Section 5 and must make sure that all objects published in digest | |
1675 | form for a particular usage and selector are published with the same | |
1676 | set of digest algorithms. | |
1677 | ||
1678 | ||
1679 | ||
1680 | Dukhovni & Hardaker Expires November 26, 2014 [Page 30] | |
1681 | \f | |
1682 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1683 | ||
1684 | ||
1685 | TLSA Publishers should follow the TLSA publication size guidance | |
1686 | found in [I-D.ietf-dane-ops] about "DANE DNS Record Size Guidelines". | |
1687 | ||
1688 | 10. Security Considerations | |
1689 | ||
1690 | This protocol leverages DANE TLSA records to implement MITM resistant | |
1691 | opportunistic channel security for SMTP. For destination domains | |
1692 | that sign their MX records and publish signed TLSA records for their | |
1693 | MX hostnames, this protocol allows sending MTAs to securely discover | |
1694 | both the availability of TLS and how to authenticate the destination. | |
1695 | ||
1696 | This protocol does not aim to secure all SMTP traffic, as that is not | |
1697 | practical until DNSSEC and DANE adoption are universal. The | |
1698 | incremental deployment provided by following this specification is a | |
1699 | best possible path for securing SMTP. This protocol coexists and | |
1700 | interoperates with the existing insecure Internet email backbone. | |
1701 | ||
1702 | The protocol does not preclude existing non-opportunistic SMTP TLS | |
1703 | security arrangements, which can continue to be used as before via | |
1704 | manual configuration with negotiated out-of-band key and TLS | |
1705 | configuration exchanges. | |
1706 | ||
1707 | Opportunistic SMTP TLS depends critically on DNSSEC for downgrade | |
1708 | resistance and secure resolution of the destination name. If DNSSEC | |
1709 | is compromised, it is not possible to fall back on the public CA PKI | |
1710 | to prevent MITM attacks. A successful breach of DNSSEC enables the | |
1711 | attacker to publish TLSA usage 3 certificate associations, and | |
1712 | thereby bypass any security benefit the legitimate domain owner might | |
1713 | hope to gain by publishing usage 0 or 1 TLSA RRs. Given the lack of | |
1714 | public CA PKI support in existing MTA deployments, avoiding | |
1715 | certificate usages 0 and 1 simplifies implementation and deployment | |
1716 | with no adverse security consequences. | |
1717 | ||
1718 | Implementations must strictly follow the portions of this | |
1719 | specification that indicate when it is appropriate to initiate a non- | |
1720 | authenticated connection or cleartext connection to a SMTP server. | |
1721 | Specifically, in order to prevent downgrade attacks on this protocol, | |
1722 | implementation must not initiate a connection when this specification | |
1723 | indicates a particular SMTP server must be considered unreachable. | |
1724 | ||
1725 | 11. IANA considerations | |
1726 | ||
1727 | This specification requires no support from IANA. | |
1728 | ||
1729 | 12. Acknowledgements | |
1730 | ||
1731 | The authors would like to extend great thanks to Tony Finch, who | |
1732 | started the original version of a DANE SMTP document. His work is | |
1733 | ||
1734 | ||
1735 | ||
1736 | Dukhovni & Hardaker Expires November 26, 2014 [Page 31] | |
1737 | \f | |
1738 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1739 | ||
1740 | ||
1741 | greatly appreciated and has been incorporated into this document. | |
1742 | The authors would like to additionally thank Phil Pennock for his | |
1743 | comments and advice on this document. | |
1744 | ||
1745 | Acknowledgments from Viktor: Thanks to Paul Hoffman who motivated me | |
1746 | to begin work on this memo and provided feedback on early drafts. | |
1747 | Thanks to Patrick Koetter, Perry Metzger and Nico Williams for | |
1748 | valuable review comments. Thanks also to Wietse Venema who created | |
1749 | Postfix, and whose advice and feedback were essential to the | |
1750 | development of the Postfix DANE implementation. | |
1751 | ||
1752 | 13. References | |
1753 | ||
1754 | 13.1. Normative References | |
1755 | ||
1756 | [I-D.ietf-dane-ops] | |
1757 | Dukhovni, V. and W. Hardaker, "DANE TLSA implementation | |
1758 | and operational guidance", draft-ietf-dane-ops-00 (work in | |
1759 | progress), October 2013. | |
1760 | ||
1761 | [RFC1035] Mockapetris, P., "Domain names - implementation and | |
1762 | specification", STD 13, RFC 1035, November 1987. | |
1763 | ||
1764 | [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate | |
1765 | Requirement Levels", BCP 14, RFC 2119, March 1997. | |
1766 | ||
1767 | [RFC3207] Hoffman, P., "SMTP Service Extension for Secure SMTP over | |
1768 | Transport Layer Security", RFC 3207, February 2002. | |
1769 | ||
1770 | [RFC4033] Arends, R., Austein, R., Larson, M., Massey, D., and S. | |
1771 | Rose, "DNS Security Introduction and Requirements", RFC | |
1772 | 4033, March 2005. | |
1773 | ||
1774 | [RFC4034] Arends, R., Austein, R., Larson, M., Massey, D., and S. | |
1775 | Rose, "Resource Records for the DNS Security Extensions", | |
1776 | RFC 4034, March 2005. | |
1777 | ||
1778 | [RFC4035] Arends, R., Austein, R., Larson, M., Massey, D., and S. | |
1779 | Rose, "Protocol Modifications for the DNS Security | |
1780 | Extensions", RFC 4035, March 2005. | |
1781 | ||
1782 | [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security | |
1783 | (TLS) Protocol Version 1.2", RFC 5246, August 2008. | |
1784 | ||
1785 | [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S., | |
1786 | Housley, R., and W. Polk, "Internet X.509 Public Key | |
1787 | Infrastructure Certificate and Certificate Revocation List | |
1788 | (CRL) Profile", RFC 5280, May 2008. | |
1789 | ||
1790 | ||
1791 | ||
1792 | Dukhovni & Hardaker Expires November 26, 2014 [Page 32] | |
1793 | \f | |
1794 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1795 | ||
1796 | ||
1797 | [RFC5321] Klensin, J., "Simple Mail Transfer Protocol", RFC 5321, | |
1798 | October 2008. | |
1799 | ||
1800 | [RFC6066] Eastlake, D., "Transport Layer Security (TLS) Extensions: | |
1801 | Extension Definitions", RFC 6066, January 2011. | |
1802 | ||
1803 | [RFC6125] Saint-Andre, P. and J. Hodges, "Representation and | |
1804 | Verification of Domain-Based Application Service Identity | |
1805 | within Internet Public Key Infrastructure Using X.509 | |
1806 | (PKIX) Certificates in the Context of Transport Layer | |
1807 | Security (TLS)", RFC 6125, March 2011. | |
1808 | ||
1809 | [RFC6186] Daboo, C., "Use of SRV Records for Locating Email | |
1810 | Submission/Access Services", RFC 6186, March 2011. | |
1811 | ||
1812 | [RFC6672] Rose, S. and W. Wijngaards, "DNAME Redirection in the | |
1813 | DNS", RFC 6672, June 2012. | |
1814 | ||
1815 | [RFC6698] Hoffman, P. and J. Schlyter, "The DNS-Based Authentication | |
1816 | of Named Entities (DANE) Transport Layer Security (TLS) | |
1817 | Protocol: TLSA", RFC 6698, August 2012. | |
1818 | ||
1819 | 13.2. Informative References | |
1820 | ||
1821 | [I-D.ietf-dane-registry-acronyms] | |
1822 | Gudmundsson, O., "Adding acronyms to simplify DANE | |
1823 | conversations", draft-ietf-dane-registry-acronyms-01 (work | |
1824 | in progress), October 2013. | |
1825 | ||
1826 | [I-D.ietf-dane-srv] | |
1827 | Finch, T., "Using DNS-Based Authentication of Named | |
1828 | Entities (DANE) TLSA records with SRV and MX records.", | |
1829 | draft-ietf-dane-srv-02 (work in progress), February 2013. | |
1830 | ||
1831 | [RFC5598] Crocker, D., "Internet Mail Architecture", RFC 5598, July | |
1832 | 2009. | |
1833 | ||
1834 | [RFC6409] Gellens, R. and J. Klensin, "Message Submission for Mail", | |
1835 | STD 72, RFC 6409, November 2011. | |
1836 | ||
1837 | Authors' Addresses | |
1838 | ||
1839 | Viktor Dukhovni | |
1840 | Two Sigma | |
1841 | ||
1842 | Email: ietf-dane@dukhovni.org | |
1843 | ||
1844 | ||
1845 | ||
1846 | ||
1847 | ||
1848 | Dukhovni & Hardaker Expires November 26, 2014 [Page 33] | |
1849 | \f | |
1850 | Internet-Draft SMTP security via opportunistic DANE TLS May 2014 | |
1851 | ||
1852 | ||
1853 | Wes Hardaker | |
1854 | Parsons | |
1855 | P.O. Box 382 | |
1856 | Davis, CA 95617 | |
1857 | US | |
1858 | ||
1859 | Email: ietf@hardakers.net | |
1860 | ||
1861 | ||
1862 | ||
1863 | ||
1864 | ||
1865 | ||
1866 | ||
1867 | ||
1868 | ||
1869 | ||
1870 | ||
1871 | ||
1872 | ||
1873 | ||
1874 | ||
1875 | ||
1876 | ||
1877 | ||
1878 | ||
1879 | ||
1880 | ||
1881 | ||
1882 | ||
1883 | ||
1884 | ||
1885 | ||
1886 | ||
1887 | ||
1888 | ||
1889 | ||
1890 | ||
1891 | ||
1892 | ||
1893 | ||
1894 | ||
1895 | ||
1896 | ||
1897 | ||
1898 | ||
1899 | ||
1900 | ||
1901 | ||
1902 | ||
1903 | ||
1904 | Dukhovni & Hardaker Expires November 26, 2014 [Page 34] |