Network Working Group M. Stecher Request for Comments: 4902 Secure Computing Category: Informational May 2007
Integrity, Privacy, and Security in Open Pluggable Edge Services (OPES) for SMTP
Status of This Memo
This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited.
Copyright (C) The IETF Trust (2007).
The Open Pluggable Edge Services (OPES) framework is application agnostic. Application-specific adaptations extend that framework. Previous work has focused on HTTP and work for SMTP is in progress. These protocols differ fundamentally in the way data flows, and it turns out that existing OPES requirements and IAB considerations for OPES need to be reviewed with regards to how well they fit for SMTP adaptation. This document analyzes aspects about the integrity of SMTP and mail message adaptation by OPES systems and about privacy and security issues when the OPES framework is adapted to SMTP. It also lists requirements that must be considered when creating the "SMTP adaptation with OPES" document.
The intent of this document is to capture this information before the current OPES working group shuts down. This is to provide input for subsequent working groups or individual contributors that may pick up the OPES/SMTP work at a later date.
Stecher Informational [Page 1]
RFC 4902 OPES/SMTP Security May 2007
Table of Contents
1. Introduction ....................................................3 1.1. Differences between Unidirectional and Bidirectional Application Protocols ........................3 1.2. Non-Standardized SMTP Adaptations at SMTP Gateways .........3 1.3. Non-OPES Issues of SMTP ....................................4 1.4. Opportunities of OPES/SMTP to Address Some Issues ..........4 1.5. Limitations of OPES in Regards to Fixing SMTP Issues .......4 2. Terminology .....................................................5 3. Integrity, Privacy, and Security Considerations .................5 3.1. Tracing Information in OPES/SMTP ...........................5 3.2. Bypass in OPES/SMTP ........................................6 3.3. Compatibility with Cryptographic Protection Mechanisms .....7 4. Protocol Requirements for OPES/SMTP .............................8 5. IAB Considerations for OPES/SMTP ................................9 5.1. IAB Consideration (2.1) One-Party Consent ..................9 5.2. IAB Consideration (2.2) IP-Layer Communications ............9 5.3. IAB Consideration (3.1) Notification .......................9 5.4. IAB Consideration (3.2) Notification ......................10 5.5. IAB Consideration (3.3) Non-Blocking ......................10 5.6. IAB Consideration Application Layer Addresses (4.x) .......10 5.7. IAB Consideration (5.1) Privacy ...........................10 5.8. IAB Consideration Encryption ..............................11 6. Security Considerations ........................................11 7. References .....................................................11 7.1. Normative References ......................................11 7.2. Informative References ....................................11 Appendix A. Acknowledgements ......................................13
Because OPES is a protocol that is built over application layer transports, its security may depend on the specifics of the transport. OPES designs are guided by the IAB considerations for OPES document , and those considerations are revisited here in the context of the SMTP protocol.
Section 3 of the OPES SMTP use cases document  maps some email and SMTP elements to OPES names that are used in this document.
1.1. Differences between Unidirectional and Bidirectional Application Protocols
The IAB listed considerations for Open Pluggable Edge Services (OPES) in  and OPES treatment of those considerations has been discussed in . Both documents make use of HTTP as an example for the underlying protocol in OPES flows, and focus on web protocols that have requests and responses in the classic form (client sends a request to a server that replies with a response of the same protocol within a single protocol transaction).
RFC 3914  already indicates that other protocols may not fit in this context, for example in Section 5.3, "Moreover, some application protocols may not have explicit responses...".
When using SMTP there are still client and server applications, and requests and responses handled within SMTP, but email messages are sent by the data provider to the recipients (data consumers) without a previous request. At that abstraction layer, email delivery via SMTP is a unidirectional process and different from the previously handled web protocols such as HTTP. For example, bypass has been defined for OPES, so far, by the data consumer requesting an OPES bypass by adding information to the application protocol request; the OPES system can then react on the bypass request in both the application request and response. For SMTP, the data consumer (email recipient) cannot request in-band that the OPES bypass handling of his/her messages.
The IAB considerations need to be revisited and special requirements may be needed for OPES handling of SMTP.
1.2. Non-Standardized SMTP Adaptations at SMTP Gateways
A large number of email filters are deployed at SMTP gateways today. In fact, all use cases listed in "OPES SMTP Use Cases"  are already deployed, often in non-standardized ways. This opens a number of integrity, privacy, and security concerns that are not
Stecher Informational [Page 3]
RFC 4902 OPES/SMTP Security May 2007
addressed, and SMTP itself does not provide effective measures to detect and defend against compromised implementations.
OPES will most likely not be able to solve these issues completely, but at least should be able to improve the situation to some extent.
The SMTP specifications  require that NDRs (Non-Delivery Reports) be sent to the originator of an undeliverable mail that has been accepted by an SMTP server. But it has become common practice for some sorts of mail (spam, worms) to be silently dropped without sending an NDR, a violation of the MUST statement of SMTP (see Section 3.7 of ). While the user of a web protocol notices if a resource cannot be fetched, neither the email sender nor email recipient may notice that an email was not delivered. These kind of issues already exist and are not introduced by OPES.
1.4. Opportunities of OPES/SMTP to Address Some Issues
Adding SMTP adaptations with OPES allows us to define a standardized way for SMTP gateway filtering, to offload filtering services to callout servers and address a number of the integrity, privacy, and security issues. OPES offers methods to add OPES tracing information and to request a bypass of filtering, and by that can make email gateway filtering a more reliable and standardized function. But OPES won't make email delivery via SMTP a reliable communication.
1.5. Limitations of OPES in Regards to Fixing SMTP Issues
The biggest concerns when adding OPES services to a network flow are that compromised, misconfigured, or faulty OPES systems may change messages in a way that the consumer can no longer read them or that messages are no longer delivered at all.
Defining a standard way to mark mails that have been handled by OPES systems is fairly simple and does not require new techniques by SMTP gateways; they already today MUST leave tracing information by adding "Received" headers to mails. Therefore, recipients receiving broken mail have a fair chance of finding the compromised OPES system by using the trace information. There is still no guarantee, as the email may have been broken in a way that makes even the tracing information unreadable. But the chance will be even better than with other protocols such as HTTP, because most email clients allow the user to display mail headers, while many browsers have no mechanism to show the HTTP headers that might include tracing info.
Stecher Informational [Page 4]
RFC 4902 OPES/SMTP Security May 2007
Email that cannot be delivered, because a compromised OPES system prevented the delivery of legitimate mail, MUST result in a an NDR to be sent to the originator of the mail according to the SMTP specifications . OPES should not be forced to fix the issue that NDRs are not reliable over SMTP.
The keywords "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in . When used with the normative meanings, these keywords will be all uppercase. Occurrences of these words in lowercase comprise normal prose usage, with no normative implications.
3. Integrity, Privacy, and Security Considerations
Tracing OPES operations is an important requirement for OPES systems. Tracing information added to email should follow a similar syntax and structure to that defined for OPES/HTTP in HTTP Adaptation with Open Pluggable Edge Services , and with the same guidelines as the SMTP specifications  defined for the "Received" headers. (We do not specify here whether "Received" headers would be used to carry the OPES information, or new trace headers should be defined, such as OPES-System and OPES-Via.)
OPES/SMTP specifications defining tracing requirements MUST be compliant with the general OPES tracing requirements defined in OPES Entities & End Points Communication , but MAY further restrict those. For example, they might require the following: that the OPES processor must add tracing information for the OPES system before calling the first callout server; that it has to augment the tracing information with additional data if necessary after the message returns from each service it is calling; and that it must ensure that the tracing information has not been deleted by an OPES service before it forwards the SMTP message.
Trace information can then be seen by mail recipients when the mail message reaches the recipient.
Mail that cannot be delivered or that is blocked by the OPES service will either be rejected or cannot be delivered after it has been accepted by an SMTP server. In the latter case, SMTP specifications  require that an NDR MUST be sent to the originator; OPES further requires that an NDR generated due to OPES processing MUST also contain information about the OPES system so that the sender gets
Stecher Informational [Page 5]
RFC 4902 OPES/SMTP Security May 2007
informed. If an email is rejected at the SMTP protocol level due to OPES processing, an OPES system MUST also include trace data in the SMTP response so that the originator can find out why and where the mail was rejected.
If a mail message was rejected or could not be delivered (and an NDR was sent), the originator of the message may want to bypass the OPES system that blocked the message.
If the recipient of a message receives a mail with OPES trace information, he may want to receive a non-OPES version of the message. Although there is no direct in-band request from the recipient back to the OPES system, the recipient can contact the sender and ask her to send the message again and to add a bypass request for the OPES system. Not all OPES systems will be allowed to fulfill a bypass request according to their policy. For example, malware scanners should not be bypassed. But other OPES services are good candidates for bypass requests, such as language translation of the email message. Translation could be bypassed after the recipient has noticed that the translated result does not meet his/her expectations and that the original message would be preferred.
An OPES system MAY also define out-of-band methods to request a bypass, for example, a web interface or an email message sent to the server that results in the creation of a white list entry for the sender/recipient pair. Examples for these out-of-band methods are email systems that keep a copy of the original email in a quarantine queue and only send the recipient a block notification, plus either a direct link or a digest notification, with the ability to retrieve the original message from quarantine. These out-of-band methods are typically offered by spam filters today.
OPES MUST implement methods to request a bypass, but there cannot be a guarantee that the bypass request will be approved. The security needs of the receiver or the receiver's network may demand that certain filters must not be bypassed (such as virus scanners). In general, the receiver should be able to configure a client centric OPES system, i.e. the receiver should be able to indicate if he/she wants to receive a non-OPES version if it is available.
Bypass requests could be added to the mail message or within the SMTP dialog. Bypass request data added to the mail message cannot bypass OPES services that operate on other SMTP dialog commands, which are sent before the mail message has been received (such as RCPT commands).
Stecher Informational [Page 6]
RFC 4902 OPES/SMTP Security May 2007
Bypass request data sent using an ESMTP extension as part of the SMTP dialog may not reach the OPES system if intermediate SMTP relays do not support those bypass request commands and don't forward that information.
3.3. Compatibility with Cryptographic Protection Mechanisms
Cryptography can be used to assure message privacy, to authenticate the originator of messages, and to detect message modification. There are standard methods for achieving some or all these protections for generic messages (, , ), and these can be used to protect SMTP data without changing the SMTP protocol.
The content of encrypted mail messages cannot be inspected by OPES systems because only the intended recipient has the information necessary for decryption. The IAB and others have suggested that users might want to share that information with OPES systems, thus permitting decryption by intermediates. For most cryptographic systems that are compatible with email, this would require end users to share their most valuable keys, in essence their "identities", with OPES machines. Some key management systems, particularly those which have centralized administrative control of keys, might have trust models in which such sharing would be sensible and secure.
After decrypting the message, an OPES box that modified the content would be faced with the task of re-encrypting it in order to maintain some semblance of "end-to-end" privacy.
If OPES/SMTP had a way to interact with end users on a per-message basis, it might be possible to communicate cryptographic key information from individual messages to end users, have them compute the message encrypting key for particular message, and to send that back to the OPES box. This would perhaps ameliorate the need to share a user's "master" message decrypting key with the OPES box. This kind of communication has not been defined for OPES.
Message protection systems generally include some message integrity mechanisms by which a recipient can check for a message modification that may have occurred after the sender released the message. This protection can be applied to encrypted or plaintext messages and can be accomplished through either symmetric or asymmetric cryptography. In the case of symmetric cryptography, the key sharing problem is exactly similar to the encryption case discussed previously. If the OPES box modified the content, then the message integrity (or authentication) code would have to be recalculated and included with the modified message.
Stecher Informational [Page 7]
RFC 4902 OPES/SMTP Security May 2007
For asymmetric cryptography the situation is more complicated. The message integrity is tied to the sender's public key, and although anyone who can get the sender's public key can also check for a message modification, no one but the sender can compute the sender's signature on a modified message. Thus, an OPES system could not modify messages and have them appear to come from the purported sender. The notion of sharing the sender's signing key with the OPES system is unpalatable because few trust models would be compatible with sharing digital identities across organization boundaries. However, if the OPES system doing the modification were under the control of the sender's local administration, the sharing might be sensible (as discussed for decryption, above).
OPES/SMTP systems could present modified content showing the modified regions in a form that permits the authentication of the original message and authentication of the OPES modifications (assuming the OPES box had a digital signature identity and key). One method for doing this is outlined in , but to our knowledge this method is not in any standard.
There are security risks associated with sharing cryptographic keys that must be addressed by implementers. Because this is not a simple task, it is not a requirement for OPES/SMTP.
The IAB recommends that all OPES services be explicitly authorized by one of the application-layer end-hosts (that is, either the data consumer application or the data provider application). For OPES/ SMTP, this means consent of either the email message sender or the recipient.
The application agnostic architecture of OPES  requires that "OPES processors MUST be consented to by either the data consumer or data provider application" (OPES processor is the email gateway for OPES/ SMTP). This cannot prevent the consent-less introduction of OPES processors by noncompliant OPES entities.
"The overall OPES framework needs to assist content providers in detecting and responding to client-centric actions by OPES intermediaries that are deemed inappropriate by the content provider" .
For OPES/SMTP this translates into assistance for the email message sender to detect and respond to recipient-centric actions that are deemed inappropriate by the sender.
This has been addressed in Section 3.1 and by the second tracing requirements in Section 4. As discussed in Section 1.3, OPES/SMTP cannot fix cases where NDRs are not sent or get blocked before reaching the sender of the original message.
"The overall OPES framework should assist end users in detecting the behavior of OPES intermediaries, potentially allowing them to identify imperfect or compromised intermediaries" .
This is addressed in Section 3.1 and by the first tracing requirement in Section 4. It must be noted that some email systems do not make the email headers available to the end user, although the headers belong to the payload that is transferred via SMTP. Building an OPES architecture with those email systems should be avoided or requires that the tracing information be made available to the end users in a different way.
"If there exists a "non-OPES" version of content available from the content provider, the OPES architecture must not prevent users from retrieving this "non-OPES" version from the content provider" .
For OPES/SMTP, this has been discussed in Section 3.2 and is addressed by the two bypass requirements of Section 4.
While "most application layer addressing revolves around URIs" (section 8 of ), SMTP uses email addresses, for which the considerations only apply to some degree.
The SMTP use cases document  includes a use case for Mail Rerouting and Address Rewriting. Alias and email list address resolution are standard functions of an email gateway described in .
Translating the reference validity consideration regarding inter- and intra-document reference validity to SMTP, OPES services mapping internal to external email addresses MUST ensure the proper mapping of addresses in all affected email headers.
"If OPES was compatible with end-to-end encryption, this would effectively ensure that OPES boxes would be restricted to ones that are known, trusted, explicitly addressed at the IP layer, and authorized (by the provision of decryption keys) by at least one of the ends" .
The document itself discusses security considerations of OPES/SMTP. General security threats of OPES are described in Security Threats for OPES 
Section 3.3 ("Compatibility with Cryptographic Protection Mechanisms") mentions that an OPES system could eventually deal with cryptographic keys. This raises security issues (such as availability and storage of cryptographic keys) that must be addressed by the implementer.
Many thanks to everybody who provided input and feedback for this document. Very special thanks to Hilarie Orman for her input and suggestions, especially for the content of Section 3.3 ("Compatibility with Cryptographic Protection Mechanisms").
Martin Stecher Secure Computing Corporation Vattmannstr. 3 33100 Paderborn Germany
This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights.
This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at firstname.lastname@example.org.
Funding for the RFC Editor function is currently provided by the Internet Society.