| Network Working Group | B. Campbell |
| Internet-Draft | Independent |
| Updates: RFC 3261, RFC 3428, RFC 4975 | R. Housley |
| (if approved) | Vigil Security |
| Intended status: Standards Track | October 30, 2017 |
| Expires: May 3, 2018 |
Securing Session Initiation Protocol (SIP) based Messaging with S/MIME
draft-campbell-sip-messaging-smime-00
Mobile messaging applications used with the Session Initiation Protocol (SIP) commonly use some combination of the SIP MESSAGE method and the Message Session Relay Protocol (MSRP). While these provide mechanisms for hop-by-hop security, neither natively provides end-to-end protection. This document offers guidance on how to provide end-to-end authentication, integrity protection, and confidentiality using the Secure/Multipurpose Internet Mail Extensions (S/MIME). It updates and provides clarifications for RFC 3261, RFC 3428, and RFC 4975.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on May 3, 2018.
Copyright (c) 2017 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Several Mobile Messaging systems use the Session Initiation Protocol (SIP) [RFC3261], typically as some combination of the SIP MESSAGE method [RFC3428] and the Message Session Relay Protocol (MSRP) [RFC4975]. For example, VoLTE uses the SIP MESSAGE method to send Short Message Service (SMS) messages. The Open Mobile Alliance (OMA) Converged IP Messaging (CPM) system uses the SIP Message Method for short “pager mode” messages and MSRP for large messages and for sessions of messages. The GSM Association (GMSA) rich communication services (RCS) uses CPM for messaging.
At the same time, organizations increasingly depend on mobile messaging systems to send notifications to their customers. Many of these notifications are security sensitive. For example, such notifications are commonly used for notice of financial transactions, notice of login or password change attempts, and sending of two-factor authentication codes.
While both SIP and MSRP provide mechanisms for hop-by-hop security, neither provides native end-to-end protection. Instead, they depend on S/MIME [RFC5750][RFC5751].
This document updates and provides clarifications to RFC 3261, RFC 3428, and RFC 4975. While each of those documents already describes the use of S/MIME to some degree, that guidance contains inconsistencies, and it is out of date in terms of supported and recommended algorithms. The guidance in RFC 3261 is offered in the context of signaling applications, and it is not entirely appropriate for messaging applications.
Both SIP and MSRP can be used to transport any content using Multipurpose Internet Mail Extensions (MIME) formats. The SIP MESSAGE method is typically limited to short messages (under 1300 octets for the MESSAGE request). MSRP can carry arbitrarily large messages, and can break large messages into chunks.
MSRP sessions are negotiated using the Session Description Protocol (SDP) [RFC4566] offer/answer mechanism [RFC3264] or similar mechanisms. This document assumes that SIP is used for the offer/answer exchange. However, the techniques should be adaptable to other signaling protocols.
The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in BCP 14 [RFC2119][RFC8174] when, and only when, they appear in all capitals, as shown here.
This document discusses the use of S/MIME with SIP based messaging. Other standardized messaging protocols exist, such as the Extensible Messaging and Presence Protocol (XMPP) [RFC6121]. Likewise, other end-to-end protection formats exist, such as JSON Web Signatures [RFC7515] and JSON Web Encryption [RFC7516].
This document focuses on SIP-based messaging because its use is becoming more common in mobile environments. It focuses on S/MIME since several mobile operating systems already have S/MIME libraries installed. While there may also be value in specifying end-to-end security for other messaging and security mechanisms, it is out of scope for this document.
This document primarily covers the sending of single messages, for example “pager-mode messages” send using the SIP MESSASGE method and “large messages” sent in MSRP. Techniques to use a common signing or encryption across a session of messages are out of scope for this document, but may be discussed in a future version.
Cryptographic algorithm requirements in this document are intended supplement those of SIP and MSRP.
The Cryptographic Message Syntax (CMS) [RFC5652] is an encapsulation syntax that is used to digitally sign, digest, authenticate, or encrypt arbitrary message content. The CMS supports a variety of architectures for certificate-based key management, especially the one defined by the IETF PKIX (Public Key Infrastructure using X.509) working group [RFC5280]. The CMS values are generated using ASN.1 [X680], using the Basic Encoding Rules (BER) and Distinguished Encoding Rules (DER) [X690].
The S/MIME Message Specification [RFC5751] defines MIME body parts based on the CMS. In this document, the application/pkcs7-mime media type is used to digitally sign an encapsulated body part, and it is also is used to encrypt an encapsulated body part.
While both SIP and MSRP require support for the multipart/signed format, this document recommends the use of application/pkcs7-mime for most signed messages. Experience with the use of S/MIME in electronic mail has shown that multipart/signed bodies are at greater risk of “helpful” tampering by intermediaries, a common cause of signature validation failure. This risk is also present for messaging applications; for example, intermediaries might insert Instant Message Delivery notification requests into messages (see Section 9.2). The application/pkcs7-mime format is also more compact, which can be important for messaging applications, especially when using the SIP MESSAGE method (see Section 7.1). The use of multipart/signed may still make sense if the message needs to be readable by receiving agents that do not support S/MIME.
When generating a signed message, sending user agents (UAs) SHOULD follow the conventions specified in [RFC5751] for the application/pkcs7-mime media type with smime-type=signed-data. When validating a signed message, receiving UAs MUST follow the conventions specified in [RFC5751] for the application/pkcs7-mime media type with smime-type=signed-data.
Sending and receiving UAs MUST support the SHA-256 message digest algorithm [RFC5754]. For convenience, the SHA-256 algorithm identifier is repeated here:
id-sha256 OBJECT IDENTIFIER ::= {
joint-iso-itu-t(2) country(16) us(840) organization(1) gov(101)
csor(3) nistalgorithm(4) hashalgs(2) 1 }
Sending and receiving UAs MAY support other message digest algorithms.
Sending and receiving UAs MUST support the Elliptic Curve Digital Signature Algorithm (ECDSA) using the NIST P256 elliptic curve and the SHA-256 message digest algorithm [RFC5480][RFC5753]. For convenience, the ECDSA with SHA-256 algorithm identifier and the object identifier for the well-known NIST P256 elliptic curve are repeated here:
ecdsa-with-SHA256 OBJECT IDENTIFIER ::= {
iso(1) member-body(2) us(840) ansi-X9-62(10045) signatures(4)
ecdsa-with-SHA2(3) 2 }
-- Note: the NIST P256 elliptic curve is also known as secp256r1.
secp256r1 OBJECT IDENTIFIER ::= {
iso(1) member-body(2) us(840) ansi-X9-62(10045) curves(3)
prime(1) 7 }
When generating an encrypted message, sending UAs MUST follow the conventions specified in [RFC5751] for the application/pkcs7-mime media type with smime-type=enveloped-data. When decrypting a received message, receiving UAs MUST follow the conventions specified in [RFC5751] for the application/pkcs7-mime media type with smime-type=enveloped-data.
Sending and receiving UAs MUST support the AES-128-CBC for content encryption [RFC3565]. For convenience, the AES-128-CBC algorithm identifier is repeated here:
id-aes128-CBC OBJECT IDENTIFIER ::= {
joint-iso-itu-t(2) country(16) us(840) organization(1) gov(101)
csor(3) nistAlgorithm(4) aes(1) 2 }
Sending and receiving UAs MAY support other content encryption algorithms.
Sending and receiving UAs MUST support the AES-128-WRAP for encryption of one AES key with another AES key [RFC3565]. For convenience, the AES-128-WRAP algorithm identifier is repeated here:
id-aes128-wrap OBJECT IDENTIFIER ::= {
joint-iso-itu-t(2) country(16) us(840) organization(1) gov(101)
csor(3) nistAlgorithm(4) aes(1) 5 }
Sending and receiving UAs MAY support other key encryption algorithms.
Symmetric key-encryption keys can be distributed before messages are sent. If sending and receiving UAs support previously distributed key-encryption keys, then they MUST assign a KEK identifier [RFC5652] to the previously distributed symmetric key.
Alternatively, a key agreement algorithm can be used to establish a single-use key-encryption key. If sending and receiving UAs support key agreement, then they MUST the Elliptic Curve Diffie-Hellman (ECDH) using the NIST P256 elliptic curve and the ANSI-X9.63-KDF key derivation function with the SHA-256 message digest algorithm [RFC5753]. For convenience, the ECDH using the ANSI-X9.63-KDF with SHA-256 algorithm identifier is repeated here:
dhSinglePass-stdDH-sha256kdf-scheme OBJECT IDENTIFIER ::= {
iso(1) identified-organization(3) certicom(132)
schemes(1) 11 1 }
When generating a signed and encrypted message, sending UAs MUST sign the message first, and then encrypt it.
Sending and receiving UAs MUST follow the S/MIME certificate handling procedures [RFC5750], with a few exceptions detailed below.
The subject alternative name extension is used as the preferred means to convey the SIP URI of a message signer. Any SIP URI present MUST be encoded using the uniformResourceIdentifier CHOICE of the GeneralName type as described in [RFC5280], Section 4.2.1.6. Since the SubjectAltName type is a SEQUENCE OF GeneralName, multiple URIs MAY be present.
When validating a certificate, receiving UAs MUST support the Elliptic Curve Digital Signature Algorithm (ECDSA) using the NIST P256 elliptic curve and the SHA-256 message digest algorithm [RFC5480].
Sending and receiving UAs MAY support other digital signature algorithms for certificate validation.
SIP and MSRP UAs are always capable of receiving binary data. Inner S/MIME entities do not require base64 encoding [RFC4648].
Both SIP and MSRP provide 8-bit safe transport channels; base64 encoding is not generally needed for the outer S/MIME entities. However, if there is a chance a message might cross a 7-bit transport (for example, gateways that convert to a 7-bit transport for intermediate transfer), base64 encoding may be needed for the outer entity.
Messaging UAs may implement a subset of S/MIME capabilities. Even when implemented, some features may not be available due to configuration. For example, UAs that do not have user certificates cannot sign messages on behalf of the user or decrypt encrypted messages sent to the user. At a minimum, a UA that supports S/MIME MUST be able to validate a signed message.
SIP and MSRP UAs advertise their level of support for S/MIME by indicating their capability to receive the “application/pkcs7-mime” media type.
The fact that a UA indicates support for the “multipart/signed” media type does not necessarily imply support for S/MIME. The UA might just be able to display clear-signed content without validating the signature. UAs that wish to indicate the ability to validate signatures for clear-signed messages MUST also indicate support for “application/pkcs7-signature”.
A UA can indicate that it can receive all smime-types by advertising “application/pkcs7-mime” with no parameters. If a UA does not accept all smime-types, it advertises the media type with the appropriate parameters. If more than one are supported, the UA includes a separate instance of the media-type string, appropriately parameterized, for each.
For example, a UA that can only received signed-data would advertise “application/pkcs7-mime; smime-type=signed-data”.
SIP signaling can fork to multiple destinations for a given Address of Record (AoR). A user might have multiple UAs with different capabilities; the capabilities remembered from an interaction with one such UA might not apply to another.
UAs can also advertise or discover S/MIME using out of band mechanisms. Such mechanisms are beyond the scope of this document.
The use of S/MIME with the SIP MESSAGE method is described in section 11.3 of [RFC3428], and for SIP in general in section 23 of [RFC3261]. This section and its child sections offer clarifications for the use of S/MIME with the SIP MESSAGE method, along with related updates to RFC 3261 and RFC 3428.
SIP MESSAGE requests are typically limited to 1300 octets. That limit applies to the entire message, including both SIP header fields and the message content. This is due to the potential for fragmentation of larger requests sent over UDP. In general, it is hard to be sure that no proxy or other intermediary will forward a SIP request over UDP somewhere along the path. Therefore, S/MIME messages sent via SIP MESSAGE should be kept as small as possible.
[RFC3261] says that a SignedData message MUST contain a certificate to be used to validate the signature. In order to reduce the message size, this document updates that to say that a SignedData message sent in a SIP MESSAGE request SHOULD contain the certificate, but MAY omit it if the sender has reason to believe that the recipient already has the certificate in its keychain, or has some other method of accessing the certificate.
SIP user agents (UA) can indicate support for S/MIME by including the appropriate media type or types in the SIP Accept header field in a response to an OPTIONS request, or in a 415 response to a SIP request that contained an unsupported media type in the body.
UAs might be able to use the user agent capabilities framework [RFC3840] to indicate support. However doing so would require the registration of one or more media feature tags with IANA.
UAs MAY use other out-of-band methods to indicate their level of support for S/MIME.
[RFC3261] requires that the recipient of a SIP request that includes a body part of an unsupported media type and a Content-Disposition header “handling” parameter of “required” return a 415 “Unsupported Media Type” response. Given that SIP MESSAGE exists for no reason other than to deliver content in the body, it is reasonable to treat the top-level body part as always required. However [RFC3428] makes no such assertion. This document updates [RFC3428] to say that a UAC that receives a SIP MESSAGE request with an unsupported media type MUST return a 415 Unsupported Media Type” response.
[RFC3261] says that if a recipient receives an S/MIME body encrypted to the wrong certificate, it MUST return a SIP 493 (Undecipherable) response, and SHOULD send a valid certificate in that response. This is not always possible in practice for SIP MESSAGE requests. The User Agent Server (UAS) may choose not to decrypt a message until the user is ready to read it. Messages may be delivered to a message store, or sent via a store-and-forward service. This document updates RFC 3261 to say that the UAS SHOULD return a SIP 493 response if it immediately attempts to decrypt the message and determines the message was encrypted to the wrong certificate. However, it MAY return a 2XX class response if decryption is deferred.
MSRP has features that interact with the use of S/MIME. In particular, the ability to send messages in chunks, the ability to send messages of unknown size, and the use of SDP to indicate media-type support create considerations for the use of S/MIME.
MSRP allows a message to be broken into “chunks” for transmission. In this context, the term “message” refers to an entire message that one user might send to another. A chunk is a fragment of that message sent in a single MSRP SEND request. All of the chunks that make up a particular message share the same Message-ID value.
The sending user agent may break a message into chunks, which the receiving user agent will reassemble to form the complete message. Intermediaries such as MSRP Relays [RFC4976] might break chunks into smaller chunks, or might reassemble chunks into larger ones; therefore the message received by the recipient may be broken into a different number of chunks than were sent by the recipient. Intermediaries might also cause chunks to be received in a different order than sent.
The sender MUST apply any S/MIME operations to the whole message prior to breaking it into chunks. Likewise, the receiver needs to reassemble the message from its chunks prior to decrypting, validating a signature, etc.
MSRP chunks are framed using an end-line. The end-line comprises seven hyphens, a 64-bit random value taken from the start line, and a continuation flag. MRSP requires the sending user agent to scan data sent in a specific chunk to be sent ensure that the end-line does not accidentally occur as part of the sent data. This scanning occurs on a chunk rather than a whole message, consequently it must occur after the sender applies any S/MIME operations.
MSRP allows a mode of operation where a UA sends some chunks of a message prior to knowing the full length of the message. For example, a sender might send streamed data over MSRP as a single message, even though it doesn’t know the full length of that data in advance. This mode is incompatible with S/MIME, since a sending UA must apply S/MIME operations to the entire message in advance of breaking it into chunks.
Therefore, when sending a message in an S/MIME format, the sender MUST include the Byte-Range header field for every chunk, including the first chunk. The Byte-Range header field MUST include the total length of the message.
A higher layer could choose to break such streamed data into a series of messages prior to applying S/MIME operation, so that each fragment appears as a distinct S/MIME separate message in MSRP. Such mechanisms are beyond the scope for this document.
A UA that supports this specification MUST explicitly include the appropriate media type or types in the “accept-types” attribute in any SDP offer or answer that proposes MSRP. It MAY indicate that it requires S/MIME wrappers for all messages by putting appropriate S/MIME media types in the “accept-types” attribute and putting all other supported media types in the “accept-wrapped-types” attribute.
For backwards compatibility, a sender MAY treat a peer that includes an asterisk (“*”) in the “accept-types” attribute as potentially supporting S/MIME. If the peer returns an MSRP 415 response to an attempt to send an S/MIME message, the sender should treat the peer as not supporting S/MIME for the duration of the session, as indicated in [RFC4975].
MSRP allows multiple reporting modes that provide different levels of feedback. If the sender includes a Failure-Report header field with a value of “no”, it will not receive failure reports. This mode should not be used carelessly, since such a sender would never see a 415 response as described above, and would have no way to learn that the recipient could not process an S/MIME body.
MSRP URIs are ephemeral. Endpoints MUST NOT use MSRP URIs to identify certificates, or insert MSRP URIs into certificate Subject Alternative Name fields. When MSRP sessions are negotiated using SIP [RFC3261], the SIP Addresses of Record (AoRs) of the peers are used instead.
Note that MSRP allows messages to be sent between peers in either direction. A given MSRP message might be sent from the SIP offerer to the SIP answer. Thus, the the sender and recipient roles may reverse between one message and another in a given session.
Successful delivery of an S/MIME message does not indicate that the recipient successfully decrypted the contents or validated a signature. Decryption and/or validation may not occur immediately on receipt, since since the recipient may not immediately view the message, and the user agent may choose not to attempt decryption or validation until the user requests it.
Likewise, successful delivery of S/MIME enveloped data does not, on its own, indicate that the recipient supports the enclosed media type. If the peer only implicitly indicated support for the enclosed media type through the use of a wildcard in the “accept-types” or “accept-wrapped types” SDP attributes, it may not decrypt the message in time to send a 415 response.
The Common Profile for Instant Messaging (CPIM) [RFC3860] defines an abstract messaging service, with the goal of creating gateways between different messaging protocols that could relay instant messages without change. The SIP MESSAGE method and MSRP were initially designed to map to the CPIM abstractions. However, at the time of this writing, CPIM compliant gateways have not been deployed. To the authors’ knowledge, no other IM protocols have been explicitly mapped to CPIM.
CPIM also defines the abstract messaging URI scheme “im:”. As of the time of this writing, the “im:” scheme is not in common use. The use of “im:” URIs as subject alternative names in certificates is for future study.
The Common Profile for Instant Messages Message Format [RFC3862] allows UAs to attach transport-neutral metadata to arbitrary MIME content. The format was designed as a canonicalization format to allow signed data to cross protocol-converting gateways without loss of metadata needed to verify the signature. While it has not typically been used for that purpose, it has been used for other metadata applications, for example, Intant Message Delivery Notifications (IMDN)[RFC5438] and MSRP Multi-party Chat [RFC7701]
Signature and encryption operations are typically applied to the entire CPIM body part, rather than to just the CPIM payload. The use of CPIM metadata fields to identify certificates or to authenticate SIP or MSRP header fields is for further study.
The Instant Message Delivery Notification (IMDN) mechanism[RFC5438] allows both endpoints and intermediary application servers to request and to generate delivery notifications. The use of S/MIME does not impact strictly end-to-end use of IMDN. IMDN recommends that devices that are capable of doing so sign delivery notifications. It further requires that delivery notifications that result from encrypted messages also be encrypted.
However, IMDN allows intermediary application servers to insert notification requests into messages, to add routing information to messages, and to act on notification requests. It also allows list servers to aggregate delivery notifications.
Such intermediaries will be unable to read end-to-end encrypted messages in order to interpret delivery notice requests. Intermediaries that insert information into end-to-end signed messages will cause the signature validation to fail.
Examples will be added in a future version of this document.
This document makes no requests of the IANA.
The security considerations from S/MIME [RFC5750][RFC5751] and elliptic curves in CMS [RFC5753] apply. The S/MIME related security considerations from SIP [RFC3261][RFC3853], SIP MESSAGE [RFC3428], and MSRP [RFC4975] apply.
This document assumes that end-entity certificate validation is provided by a chain of trust to certification authority (CA), using a public key infrastructure. The security considerations from [RFC5280] apply. However, other validations methods may be possible; for example sending a signed fingerprint for the end-entity in SDP. The relationship of this work and the techniques discussed in [RFC4474], [I-D.ietf-stir-rfc4474bis], and [I-D.ietf-sipbrandy-rtpsec] are for further study.
When matching an end-entity certificate to the sender or recipient identity, the respective SIP AoRs are used. Typically these will match the SIP From and To header fields. Matching SIP AoRs from other header fields, for example, P-Asserted-Identity [RFC3325], is for further study.
The secure notification use case discussed in Section 1 has significant vulnerabilities when used in an insecure environment. For example, “phishing” messages could be used to trick users into revealing credentials. Eavesdroppers could learn confirmation codes from unprotected two-factor authentication messages. Unsolicited messages sent by impersonators could tarnish the reputation of an organization. While hop-by-hop protection can mitigate some of those risks, it still leaves messages vulnerabile to malicious or compromised intermediaries.
Mobile messaging is typically an online application; online certificate revocation checks should usually be feasible.