SIP Identity using Media Path
draft-wing-rtcweb-identity-media-00

Abstract

This document defines a new SIP identity mechanism which creates a signature over a certain subset of SIP headers and certain subset of SDP lines. This this mechanism works with trickling ICE candidates and works across zero or more Session Border Controllers.

Status of this Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on September 04, 2012.

Copyright Notice

Copyright (c) 2012 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.

Table of Contents

• 1. Introduction
• 2. Notational Conventions
• 3. Operation
• 3.1. Identity Media Signature
• 3.2. Authentication Service
• 3.3. Validation
• 4. Proof of Identity Techniques
• 4.1. TLS
• 4.2. DTLS-SRTP
• 5. ABNF
• 6. Security Considerations
• 6.1. Device Disclosure
• 6.2. Modification of SDP
• 7. Operational Differences from RFC4474
• 8. Limitations
• 9. Examples
• 9.1. DTLS
• 10. Acknowledgements
• 11. IANA Considerations
• 12. References
• Authors' Addresses

1. Introduction

SIP Identity [RFC4474] provides cryptographic identity for SIP requests. It provides this protection by signing certain SIP header fields (Contact, Date, Call-ID, CSeq, To, and From) and the SIP message body. This mechanism suffers from two problems. First, it is inefficient when ICE candidates are trickled (as each update to the ICE candidate list would have to be signed and validated). Second, it breaks entirely if IP addresses are modified during SIP routing, such as by a Session Border Controller.

To avoid these problems, a new mechanism is described in this document which provides cryptographic assurance of the endpoint's identity that works with ICE candidate trickling and works through most B2BUAs and through most SBCs.

The mechanism described in this document signs only certain SDP attributes (rather than all SDP attributes) and certain SIP headers. The remote endpoint is expected to validate the signature and initiate a cryptographic handshake over the media path, which proves the session has been established with the "From:" party in the SIP header. A mechanism to perform the handshake over the media path is shown using DTLS-SRTP and TLS. This mechanism is extensible so that techniques other than DTLS-SRTP or TLS can be used.

Readers of this document are expected to be familiar with "Enhancements for Authenticated Identity Management in the Session Initiation Protocol (SIP)" [RFC4474], which defines the Identity and Identity-Info header fields that are also used by this document.

2. Notational Conventions

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

3. Operation

This document can utilize an authentication proxy, a concept originally introduced in [RFC4474]. The basic steps are:

• A new header, Identity-Media, is created containing the names of certain SDP attributes from SDP bodyparts, and containing a hash of non-SDP bodyparts.
• Several SIP headers and the Identity-Media header are all signed (as detailed in Section 3.1), and the result is placed in Identity-Media-Signature.
• The receiving domain validates the signature, and if the request is an invitation to establish a media channel, performs a proof of identity validation using DTLS.

The following figure shows how the Authentication Service and the media validation is performed. The figure assumes the endpoints themselves perform the media validation.

                        :  Service   :
       Enterprise-A     : Provider-1 :    Enterprise-B
                        :            :
                 Auth.  :  B2BUA or  :  Auth.    
    Endpoint-A  Service :    SBC     : Service  Endpoint-B
        |          |    :     |      :   |         |
 1.     |--INVITE->|    :     |      :   |         |
 2.     |        sign   :     |      :   |         |
 3.     |          |-INVITE-->|-INVITE-->|         |
 4.     |          |    :     |      : validate    |
 5.     |          |    :     |      :   |-------->|
 6.     |<============= DTLS =====================>|
 7.     |          |    :     |      :   |     validated
 8.     |          |    :     |      :   |     ring phone
        |          |    :     |      :   |         |
                        :            :

Step 1 :

Originating endpoint prepares to send an INVITE and chooses the identity-challenge technique it supports, and indicates that in the SDP it generates. Described in this document are identity challenges for DTLS. It then sends the INVITE to its local SIP proxy.

Step 2 :

Originating endpoint's authentication service creates a new header, Identity-Media, containing certain attribute names from the SDP (e.g., "a=fingerprint", "a=ice-pub-key"). The authentication service then creates a signature over certain SIP headers (e.g., From, To) and this new Identity-Media header. The resulting signature is inserted into the new Identity-Media-Signature header. An Identity-Info header is added, pointing to this domain's certificate. The INVITE, with these additional headers, is forwarded to the next administrative domain.
[NOTE: alternatively, we could allow the UAC to create the Identity-Media header with the attributes it wants signed, then have the authentication server sign them and insert the signature header - this would be more flexible]

Step 3 :

The next administrative domain has an SBC (or B2BUA). The SBC modifies or rewrites certain SDP fields. Most typically an SBC will modify the "m" and "c" lines. These modifications do not break the signature, so long as the SBC doesn't remove the headers Identity-Media, Identity-Media-Signature, or Identity-Info, and do not remove or alter the signed attributes from the SDP.

Step 4 :

The terminating endpoint's authentication service receives the INVITE. It validates that the signature contained in the Identity-Media-Signature header, and validates that the signing certificate is owned by the originating domain from step 2. This validation is done by using the certificate pointed to in the Identity-Info header, which MUST match the domain in the From: address.

Step 5 :

If the validation was successful, the terminating endpoint's authentication service forwards the INVITE to the endpoint.

Step 6 :

The terminating endpoint chooses a compatible identity-challenge technique from the INVITE (TLS or DTLS-SRTP) and performs that challenge. Described in this document are identity challenges for TLS and DTLS-SRTP.

Step 7 :

The TLS and DTLS-SRTP identity challenges cause the exchange of a certificate on the media path. The terminating endpoint compares the certificate or public key with the fingerprint in the (signed) Identity-Media header (originally created in step 2). If they match, the terminating endpoint completes the identity challenge exchange. After completion, the originating endpoint has proven (to the terminating endpoint) that the originating endpoint knows the private key associated with the certificate (or public key) signed in step 2. The terminating endpoint has now validated the identity of the originating endpoint.

Step 8 :

The terminating endpoint can reliably and honestly indicate calling party information ("caller-id") and ring the phone.

3.1. Identity Media Signature

In RFC4474, a signature is formed over some SIP headers and over the entire body (which most typically contains SDP). In this specification, some SIP headers are signed but only specific SDP attributes that provide cryptographic identity are signed (e.g., "a=fingerprint" and its value). The specific SDP attributes that are signed depends on which cryptographic identity technique(s) is used; see section Section 4.

The SIP headers that are signed, for the signature placed into the Identity-Media-Signature header are:

• The AoR of the UA sending the message, or addr-spec of the From header field (referred to occasionally here as the 'identity field').
• The addr-spec component of the To header field, which is the AoR to which the request is being sent.
• The SIP method.
• [NOTE: Contact, CSeq and Call-Id not included]
• The Date header field, with exactly one space each for each SP and the weekday and month items case set as shown in the BNF in RFC3261. RFC3261 specifies that the BNF for weekday and month is a choice amongst a set of tokens. The RFC2234 rules for the BNF specify that tokens are case sensitive. However, when used to construct the canonical string defined here, the first letter of each week and month MUST be capitalized, and the remaining two letters must be lowercase. This matches the capitalization provided in the definition of each token. All requests that use the Identity-Media mechanism MUST contain a Date header.
• The Identity-Media header field value.

The hash is formed of these elements:

   digest-string = addr-spec "|" addr-spec "|"
                   Method "|" SIP-date "|"
                   attrib-bodyhash-list

The first addr-spec MUST be taken from the From header field value, the second addr-spec MUST be taken from the To header field value.

The Identity-Info header points to where the authentication service's certificate can be retrieved from.

3.2. Authentication Service

The authentication service examines the SIP message body to build the Identity-Media header. For each message body found, in the order found:

• if the body part is application/sdp, the authentication service retrieves only the cryptographic attributes from the SDP (as described in Section 4), and appends that information to the Identity-Media header.
• otherwise, for all other body parts, the body part is hashed using SHA-1, and the first 96 bytes are appended to the Identity-Media header using "BPH=".

For example, A SIP request with three bodyparts: text/plain, application/sdp, and image/jpg, the Identity-Media attribute would contain a bodypart hash of the text/plain part, certain SDP attribute lines from the application/sdp bodypart (a=fingerprint in this example), and a bodypart hash of the image/jpg bodypart:

  Identity-Media: BPH="e32je3j23cjek3dz","a=fingerprint",
    BPH="8fj289r3i892381c"
 

This Identity-Media header, along with the headers and portions of headers described in Section 3.1 are all signed by the authentication service. The resulting signature is placed on the new Identity-Media-Signature header.

3.3. Validation

The validation service can be performed by the remote endpoint itself or by a device acting on behalf of the endpoint. The validation service first checks the signature in the Identity-Media-Signature field. If this is valid, the endpoint (or its validation service operating on its behalf) then initiates a TLS or DTLS-SRTP identity proof (Section 4). This causes the originating endpoint to prove possession of its private key that corresponds to the certificate that was signed by the remote domain's authentication service.

4. Proof of Identity Techniques

Two technique is described below -- TLS and DTLS-SRTP. Previous versions of this document had described other techniques (ICE, HIP, and ZRTP). Both TLS and DTLS-SRTP cryptographically prove the identity signed by the authentication service in SIP is the same as the identity on the media path.

The authentication service creates a new Identity-Media header and places into that header those SDP attribute names associated with that technique. The authentication service then creates a signature over specific SIP headers (see Section 3.1), and places that signature into the new Identity-Media-Signature header. The SIP request is then sent outside of the originating domain.

The receiving domain validates the Identity-Media-Signature. If successful, the SIP request is forwarded to the end system. The end system initiates a DTLS session and validates that the (signed) certificate fingerprint presented in the SIP signaling matches the certificate presented in the DTLS exchange. If they match, and the DTLS exchange completes successfully, the local endpoint has validated the identity of the remote endpoint.

Note: Due to SIP forking, the calling party may receive many identity challenges, each incurring a public key operation to prove identity. Mechanisms to deal with this are for future study.

4.1. TLS

TLS uses the "fingerprint" attribute to provide a hash of the certificate in the SDP. The fingerprint attribute is defined by [RFC4572] for TLS. This a=fingerprint line is included in the Identity-Media SDP attribute.

4.2. DTLS-SRTP

DTLS uses the same "fingerprint" attribute originally described for TLS, and handled identically to TLS.

5. ABNF

The following figure shows the syntax of the new SIP header fields using ABNF [RFC5234]

   
   identity-media        = "Identity-Media" HCOLON 
                           attrib-bodyhash-list
   attrib-bodyhash-list  = attrib-bodyhash *(COMMA attrib-bodyhash)
   attrib-bodyhash       = quoted-attrib | quoted-bodyparthash
   quoted-attribute      = DQUOTE attribute DQUOTE  ; SDP "a=" line
   quoted-bodyhash       = "BPH" EQUAL DQUOTE bodyparthash DQUOTE
   bodyparthash          = 32HEXDIG

   identity-media-sig    = "Identity-Media-Signature" HCOLON 
                           signature
   signature             = DQUOT 32HEXDIG DQUOT

   Identity-Info = "Identity-Info" HCOLON ident-info
                    *( SEMI ident-info-params )
   ident-info = LAQUOT absoluteURI RAQUOT
   ident-info-params = ident-info-alg / ident-info-extension
   ident-info-alg = "alg" EQUAL token
   ident-info-extension = generic-param

6. Security Considerations

[[some of RFC4474's security considerations also apply.]]

6.1. Device Disclosure

Although the mechanism described in this paper allows SBCs to be used with a cryptographic identity scheme, it does expose the identity of the user's certificate -- which is exposed by DTLS-SRTP itself. If a unique certificate is installed on each user's device, the remote party will be able to discern which device is terminating the call. This problem is more pronounced when SIP retargeting occurs in conjunction with Connected Identity [RFC4916].

If this isn't desired, there are two solutions:

• All devices under the control of the user will need to have the same certificate (and associated private key) installed on them.
• The device needs to manufacture a new self-signed certificate (or public key) for each call, and populate the appropriate SDP attributes with that certificate (or public key). This is possible because the identity service described in this paper does not require the same certificate or public key to be used on every call.

6.2. Modification of SDP

One issue with only signing specific SDP attributes is that a man in the middle can modify the un-signed SDP for nefarious purposes, beyond simply changing m=/c= lines. In particular, an attacker could set the c= connection line used for DTLS-SRTP fingerprint to 0.0.0.0 and the m= media line to port 0, essentially disabling that offered media session. The attacker could also add a set of c=/m= lines for non-SRTP media, and thus make a non-SRTP offer with a perfectly valid identity signature. Or an attacker could insert SDP capability negotiation attributes to create a best-effort type SRTP offer, with SRTP (rather than RTP) being the lowest preference.

This draft prevents such downgrade attacks by requiring the called UA use DTLS-SRTP, HIP, ICE, or TLS on the media path to establish identity. Thus, an attacker performing the attacks described above will not successfully fool the called UA because the (intended) victim will use DTLS-SRTP (or HIP, ICE, or TLS) on the media path, and the attacker does not possess the private key of the legitimate caller.

7. Operational Differences from RFC4474

RFC4474 imposes one public key operation for the authentication service and one for validation. If Connected Identity [RFC4916] is used, only one additional public key operation is necessary for the header signature validation; the expense of the DTLS, TLS, or ICE public key operation has already been incurred by both parties and is not repeated.

RFC4474 includes the Contact URI in the signed headers. That is not required by this mechanism because it adds no security property, and will fail validation when crossing SBCs and B2BUA's. It is of dubious security value because Via/Record-Route can be inserted for response interception regardless, and some requests don't contain a Contact anyway (e.g., MESSAGE). It does not provide any replay/copy-paste protection either, for the same reasons.

RFC4474 includes the CSeq in the signed headers. That is not required by this mechanism because it adds little security, and will fail validation when crossing SBCs and B2BUA's in some cases. It is of little security value because it provides no protection from cut-paste attack for different targets, and although it would prevent replay attack within the same session, since the media key-related SDP portions are signed anyway, replaying the request will not do anything useful.

RFC4474 includes the Call-Id in the signed headers. That is not required by this mechanism because it adds little security, and will fail validation when crossing SBCs and B2BUA's in some cases. It is of little security value because it provides no protection from cut-paste attack for different targets, and although it would prevent replay attack for the same target, since the media key-related SDP portions are signed anyway, replaying the request will not do anything useful.

The mechanism described in this document has the following advantages over RFC4474:

• Only the edge network needs to create signatures on SIP requests -- not every intervening SBC,
• The original cryptographically-provable identity is preserved across any number of SBCs, B2BUA's, etc.
• SBCs, B2BUA's, and other "middle-boxes" in intermediate domains do not need to be upgraded or changed in order for the originating and terminating domains to use this new mechanism.

8. Limitations

For the identity procedure described in this document to function, every device -- including Session Border Controllers -- on the path MUST permit DTLS-SRTP on the media path. Further, those devices MUST NOT interfere with the signed SDP attributes or the new SIP headers necessary for Identity Media to operate.

For the technique described in this document to function, all on-path SIP elements -- SBCs, B2BUAs, and SIP proxies -- MUST NOT interfere with the signed headers. The identity mechanism described in this document is not harmed if on-path SIP elements alter the SDP (e.g., by deleting non-signed attributes, connection addresses, etc.).

9. Examples

9.1. DTLS

This example shows how two a=fingerprint lines in SDP would populate the Identity-Media SIP header field. The following is an example of an INVITE created by the endpoint.

(lines folded for readability)

   INVITE sip:bob@biloxi.example.org SIP/2.0
   Via: SIP/2.0/TLS pc33.atlanta.example.com;branch=z9hG4bKnashds8
   To: Bob <sip:bob@biloxi.example.org>
   From: Alice <sip:alice@atlanta.example.com>;tag=1928301774
   Call-ID: a84b4c76e66710
   CSeq: 314159 INVITE
   Max-Forwards: 70
   Date: Thu, 21 Feb 2002 13:02:03 GMT
   Contact: <sip:alice@pc33.atlanta.example.com>
   Content-Type: application/sdp
   Content-Length: 147

   v=0
   o=- 6418913922105372816 2105372818 IN IP4 192.0.2.1
   s=example2
   c=IN IP4 192.0.2.1
   t=0 0
   m=audio 54113 RTP/SAVP 0
   a=fingerprint:SHA-1 
     4A:AD:B9:B1:3F:82:18:3B:54:02:12:DF:3E:5D:49:6B:19:E5:7C:AB
   m=video 54115 RTP/SAVP 0
   a=fingerprint:SHA-1 
     4A:AD:B9:B1:3F:82:18:3B:54:02:12:DF:3E:5D:49:6B:19:E5:7C:AB

The SIP proxy performing the Media Identity authentication service would then insert the following three SIP headers into the message. The Identity-Media header contains all of the SDP attribute lines that are signed and the Identity-Media header contains the signature of all of the relevant SIP headers and of the Identity-Media header. Lines are folded for readability:

  Identity-Info: <https://atlanta.example.com/atlanta.cer>
     ;alg=rsa-sha1  
  Identity-Media: "a=fingerprint","a=fingerprint"
  Identity-Media-Signature: 
   "ZYNBbHC00VMZr2kZt6VmCvPonWJMGvQTBDqghoWeLxJfzB2a1pxAr3VgrB0SsSAa
    ifsRdiOPoQZYOy2wrVghuhcsMbHWUSFxI6p6q5TOQXHMmz6uEo3svJsSH49thyGn
    FVcnyaZ++yRlBYYQTLqWzJ+KVhPKbfU/pryhVn9Yc6U="

10. Acknowledgements

The mechanism described in this paper is derived from Jon Peterson and Cullen Jennings' [RFC4474], which was formerly a document of the SIP working group.

Thanks to Hans Persson for his suggestions which improved this document.

11. IANA Considerations

[[This section will be completed in a later version of this document.]]

12. References

Authors' Addresses