A Document Format for Expressing Authorization Policies to tackle Spam and Unwanted Communication for Internet Telephony

The problem of SPAM for Internet Telephony (SPIT) is an imminent challenge and only the combination of several techniques can provide a framework for dealing with unwanted communication, as stated in . One important building block is to have a mechanism that can instruct SIP intermediaries to react differently on incoming requests based on policies. Different entities, such as end users, parents on behalf of their children, system administrators in enterprise networks, etc., might create and modify authorization policies. The conditions in these policies can be created from many sources but some information elements are more important than others. For example, there is reason to believe that applying authorization policies based on the authenticated identity is an effective way to accept a communication attempt to deal with unsolicited communication. Authentication based on the SIP identity mechanism, see , is one important concept. The requirements for the authorization policies described in this document are outlined in . A framework document is available at .

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 RFC 2119 . This document reuses the terminology from RFC 4745 : The RM is an entity that creates the authorization policies that react to unwanted connection attempts. The rule maker might be an end user that owns the device, a VoIP service provider, a person with a relationship to the end user (e.g., the parents of a child using a mobile phone). A standardized policy language is needed when the creation, modification and deletion of authorization policies are not only a local matter. An authorization policy is given by a rule set. A rule set contains an unordered list of rules. Each rule has a condition, an action and a transformation component. The terms 'authorization policy', 'policy', 'rule set', 'authorization policy rule', 'policy rule' and 'rule' are used interchangeably. Authorization policies can be applied at the end host and/or by intermediaries. The term permission refers to the action and transformation components of a rule. We use the term 'Recipient' for the entity that is target of the communication attempt of a sender.

A SPIT authorization document is an XML document, formatted according to the schema defined in RFC 4745 . SPIT authorization documents inherit the MIME type of common policy documents, application/auth-policy+xml. As described in , this document is composed of rules which contain three parts - conditions, actions, and transformations. Each action or transformation, which is also called a permission, has the property of being a positive grant to the authorization server to perform the resulting actions, be it allow, block etc . As a result, there is a well-defined mechanism for combining actions and transformations obtained from several sources. This mechanism therefore can be used to filter connection attempts thus leading to effective SPIT prevention.

Policies are XML documents that are stored at a Proxy Server or a dedicated device. The Rule Maker therefore needs to use a protocol to create, modify and delete the authorization policies defined in this document. Such a protocol is available with the Extensible Markup Language (XML) Configuration Access Protocol (XCAP) .

This section describes the additional enhancements of the conditions-part of the rule. This document inherits the Common Policy functionality, including <identity>, <validity>, and <sphere> conditions. Note that, as discussed in , a permission document applies to a translation if all the expressions in its conditions part evaluate to TRUE.

Although the <identity> element is defined in , that specification indicates that the specific usages of the framework document need to define details that are protocol and usage specific. In particular, it is necessary for a usage of the common policy framework to: Define acceptable means of authentication. Define the procedure for representing the identity as a URI or IRI . This sub-section defines those details for systems based on .

When used with SIP, a request is considered authenticated if one of the following techniques is used: The proxy has authenticated the sender using SIP digest authentication . However, if the anonymous authentication described on page 194 of RFC 3261 was used, the sender is not considered authenticated. If a request contains a P-Asserted-ID header field and the request is coming from a trusted element, the sender is considered authenticated. If a request contains an Identity header field as defined in , and it validates the From header field of the request, the request is considered to be authenticated. Note that this is true even if the request contained a From header field of the form sip:anonymous@example.com. As long as the signature verifies that the request legitimately came from this identity, it is considered authenticated. An anonymous From header field with RFC 4474 is considered authenticated, while anonymous digest is not considered authenticated, because the former still involves the usage of an actual username and credential as part of an authentication operation in the originating domain.

For messages that are authenticated using SIP Digest, the identity of the sender is set equal to the address of record (AoR) for the user that has authenticated themselves. The AoR is always a URI, and can be either a SIP URI or tel URI . For example, consider the following "user record" in a database:

If the proxy server receives an INVITE, challenges it with the realm set to "example.com", and the subsequent INVITE contains an Authorization header field with a username of "ali" and a digest response generated with the password "f779ajvvh8a6s6", the identity used in matching operations is "sip:alice@example.com". For messages that are authenticated using RFC 3325 , the identity of the sender is equal to the URI in the P-Asserted-ID header field. If there are multiple values for the P-Asserted-ID header field (there can be one sip URI and one tel URI ), then each of them is used for the comparisons outlined in , and if either of them match a <one> or <except> element, it is considered a match. For messages that are authenticated using the SIP Identity mechanism , identity of the sender is equal to the SIP URI in the From header field of the request, assuming that the signature in the Identity header field has been validated. In SIP systems, it is possible for a user to have aliases - that is, there are multiple SIP AoRs "assigned" to a single user. In terms of this specification, there is no relationship between those aliases. Each would look like a different user. This will be the consequence for systems where the sender is in a different domain than the recipient. However, even if the sender and recipient are in the same domain, and the proxy server knows that there are aliases for the sender, these aliases are not mapped to each other or used in any way. SIP also allows for anonymous identities. If a message is anonymous because the digest challenge/response used the "anonymous" username, the message is considered unauthenticated and will match only an empty <identity> element. If a message is anonymous because it contains a Privacy header field , but still contains a P-Asserted-ID header field, the identity in the P-Asserted-ID header field is still used in the authorization computations; the fact that the message was anonymous has no impact on the identity processing. However, if the message had traversed a trust boundary and the P-Asserted-ID header field and the Privacy header field had been removed, the message will be considered unauthenticated when it arrives at the proxy server. Finally, if a message contained an Identity header field that was validated, and the From header field contained a URI of the form sip:anonymous@example.com, then the sender is considered authenticated, and it will have an identity equal to sip:anonymous@example.com. Had such an identity been placed into a <one> or <except> element, there will be a match. It is important to note that SIP frequently uses both SIP URI and tel URI as identifiers, and to make matters more confusing, a SIP URI can contain a phone number in its user part, in the same format used in a tel URI. The sender's identity that is a SIP URI with a phone number will not match the <one> and <except> conditions whose 'id' is a tel URI with the same number. The same is true in the reverse. If the sender's identity is a tel URI, this will not match a SIP URI in the <one> or <except> conditions whose user part is a phone number. URIs of different schemes are never equivalent.

The <sphere> element is defined in . However, each application making use of the common policy specification needs to determine how the policy server computes the value of the sphere to be used in the evaluation of the condition. To compute the value of <sphere>, the proxy server interacts with a presence server who knows whether at least one of the published presence documents includes the <sphere> element as part of the person data component , and all of those containing the element have the same value for it, that is the value used for the sphere in policy policy processing. If, however, the <sphere> element was not available to the presence server (and hence not for the proxy server), or it was present but had inconsistent values, its value is considered undefined in terms of policy processing.

The <spit-handling> element is a way to react on the execution of certain SPIT handling mechanisms. For example, a rule might indicate that a CAPTCHA has to be sent to the sender and the sender subsequently has to return the result. Depending on the outcome of the robot test the rules might enforce different actions. This element provides such a condition capability. The <spit-handling> condition evaluates to TRUE if any of its child elements evaluate to TRUE, i.e., the results of the individual child element are combined using a logical OR. The <spit-handling> element MAY contain zero or more <challenge> elements. The <challenge> elements has an attribute 'result' that either contains "SUCCESS" or "FAILURE".

This condition evaluates to TRUE when the called user's current presence activity status is equal to the value in the <presence-status> element. Otherwise the condition evaluates to FALSE.

The <time-period> element allows to make decisions based on the time, date and timezone. It defines an extended version of the <validity> element. The <time-period> element may contain the following attributes: Start of interval (RFC 2445 DATE-TIME). This attribute is MANDATORY. End of interval (RFC 2445 DATE-TIME). This attribute is MANDATORY. Start of time interval in a particular day. It is of the TIME data type as mentioned in Section 4.3.12 of RFC 2445 . This attribute is OPTIONAL. The default value is 000000. End of time interval in a particular day. It is of the TIME data type as mentioned in Section 4.3.12 of RFC 2445 . This attribute is OPTIONAL. The default value is 235959. List of days of the week. This attribute is OPTIONAL. The <time-period> is based on the description in CPL but with a reduced feature set. The "dtstart" and "dtend" attributes are formatted as iCalendar COS DATE-TIME values, as specified in Section 4.3.5 of RFC 2445 . Only floating or UTC times can be used with time zones. The DATE-TIME is a subset of the corresponding syntaxes from ISO 8601 . The "timestart" specifes a time value to indicate the beginning of every day. The default value is 000000 representing the beginning of the day. The "timeend" specifes a time value to indicate the end of every day. The default value is 235959 representing the end of the day. The "byweekday" attribute specifies a comma-separated list of days of the week. "MO" indicates Monday, "TU" indicates Tuesday, "WE" indicates Wednesday, "TH" indicates Thursday, "FR" indicates Friday, "SA" indicates Saturday, and "SU" indicates Sunday. These values are not case-sensitive. Here is an example of the time-period element.

]]> The following aspects need to be considered: By default, if all the OPTIONAL parameters are missing, <time-period> element is valid for the whole duration from 'dtstart' to 'dtend'. The 'byweekday' attribute comes into effect only if the period from 'dtstart' till 'dtstart' is long enough to accommodate the specified values, else they are just neglected. If the values of the 'byweekday' attribute values do not correspond to the expected domain, they are simply ignored. Only a single 'byweekday' attribute MUST be listed in a <time> element.

As stated in , conditions are the 'if'-part of rules, whereas actions and transformations form their 'then'-part. The actions and transformations parts of a rule determine which operations the proxy server MUST execute on receiving a connection request attempt that matches all conditions of this rule. Actions and transformations permit certain operations to be executed.

The <handling> element allows a couple of actions to be triggered, namely The block action states that this specific connection request MUST NOT be forwarded and a "403" forbidden message MUST be sent to the sender of the message. The Allow action states that this specific connection request MUST be forwarded. Furthermore, a couple of further mechanisms, such as computational puzzles mechanism (described in ), the consent framework (described in ) etc. can be executed. Each mechanism needs to register a URI and the value of URI is placed in this field. [Editior's Note: For editorial purposes the schema currently lists a few examples but in a non-URI format. When solution documents define these URIs then they can be used with this document.]

The action supported in this section is forwarding of calls with the <forward-to> element that contains the following child element <target> that specifies the address of the forwarding rule. It should be a valid SIP URI (RFC 3261 ) or TEL URI (RFC 3966 ).

This section provides a few examples for policy rules defined in this document.

The following policy shows a white list with an identity condition and a simple time-based condition.

2003-12-24T17:00:00+01:00 2003-12-24T19:00:00+01:00 allow ]]>

The following policy shows the usage of the <time-period> element to forward calls to an answering machine during the night.

sip:answering-machine@home.foo-bar.com

]]>

The following example policy shows three rules with the rule id r1 - r4. Rule r1 matches for authenticated identities from the domain "example.com", "example.org" and the single identity "sip:bob@good.example.net". For these conditions SIP messages are forwarded to the SIP UA as indicated with the <handling> element. Rule r2 indicates that for SIP messages where the identity has not been verifiable the hash cash mechanism and CAPTCHAs are applied (see the 'hashcash' and the 'captcha' token in the <execute> element). Rule r3 contains the <spit-handling> element with the <challenge> child element. This rule evaluates to TRUE if the sender returned a valid hash cash or a valid CAPTCHA result. The action part of the rule indicates that the call is then forwarded to the answering machine, namely sip:answering-machine@home.foo-bar.com. Rule r4 blocks the call if sender provided a wrong hash cash or CAPTCHA result. Rule r1 and r2 are valid only from 2007-01-01T01:00:00+01:00 to 2007-07-01T24:00:00+01:00.

2007-01-01T01:00:00+01:00 2007-07-01T24:00:00+01:00 allow 2007-01-01T01:00:00+01:00 2007-07-01T24:00:00+01:00 hashcash captcha

hashcash captcha

sip:answering-machine@home.foo-bar.com

hashcash captcha

block ]]>

This section contains the XML schema that defines the policies schema described in this document. This schema extends the Common Policy schema (see ) by introducing new members of the <condition> and <action> elements.

Exactly one of the two attributes "dtend" and "duration" must occur. None of the attributes following freq are meaningful unless freq appears. RFC 2445 DATE-TIME RFC 2445 DATE-TIME RFC 2445 TIME. It represents time in hours, minutes and seconds and denotes the beginning of the day time. The default value is 000000, denoting the beginning of the day. RFC 2445 TIME. It represents time in hours, minutes and seconds and denotes the end of the day time. The default value is 235959, denoting the end of the day. Comma-separated list of days of the week. Valid values are "MO", "TU", "WE", "TH", "FR", "SA" and "SU". These values are not case-sensitive. Each can be preceded by a positive (+n) or negative (-n) integer. ]]>

The following section defines the details necessary for clients to manipulate SPIT authorization documents from a server using XCAP.

XCAP requires application usages to define a unique application usage ID (AUID) in either the IETF tree or a vendor tree. This specification defines the "Spit-policy" AUID within the IETF tree, via the IANA registration in .

XCAP requires application usages to define a schema for their documents. The schema for Anti-SPIT authorization documents is described in .

XCAP requires application usages to define the default namespace for their documents. The default namespace is urn:ietf:params:xml:ns:spit-policy.

XCAP requires application usages to defined the MIME type for documents they carry. Anti-SPIT privacy authorization documents inherit the MIME type of Common Policy documents, application/auth-policy+xml.

This specification does not define additional constraints.

This document discusses the semantics of Anti-SPIT authorization.

When a SIP Proxy receives a SIP message to route it towards to a specific user foo, it will look for all documents within http://[xcaproot]/spit-policy/users/foo, and use all documents found beneath that point to guide authorization policy.

This application usage does not define additional resource interdependencies.

This application usage does not modify the default XCAP authorization policy, which is that only a user can read, write or modify his/her own documents. A server can allow privileged users to modify documents that they do not own, but the establishment and indication of such policies is outside the scope of this document.

There are several IANA considerations associated with this specification.

urn:ietf:params:xml:schema:spit-policy Hannes Tschofenig (hannes.tschofenig@nsn.com). The XML schema to be registered is contained in . Its first line is

]]> and its last line is

]]>

urn:ietf:params:xml:ns:spit-policy Hannes Tschofenig (hannes.tschofenig@nsn.com).

This section registers an XCAP Application Usage ID (AUID) according to the IANA procedures defined in . Name of the AUID: spit-policy Description: The rules defined in this documents describe ways to react on unwanted and unsolicted communication (including Spam).

This document aims to make it simple for users to influence the behavior of SIP message routing with an emphasis on SPIT prevention. This document proposes a strawman proposal for conditions and actions that might be useful when it comes to allowing a UA to tell its proxies which messages it wants to receive and what tasks it wants those proxies to perform before sending a SIP request to the UA. A couple of requirements are described in and a general discussion about the available solution mechanisms is available with . This document offers the ability to glue the different solution pieces together. Since this document uses the Common Policy framework it also inherits its capabilities, including the combining permission algorithm that is applied when multiple rules fire. Unauthorized access to the user's Anti-SPIT rules must be prevented to avoid the introduction of security vulnerabilities.

We would like to thank Mayutan Arumaithurai (mayutan.arumaithurai@gmail.com) for his work on this document.

We would like to thank Jonathan Rosenberg, David Schwartz and Dan York for sharing their thoughts with us before the first version of this document was written. Miguel Garcia and Rémi Denis-Courmont for their review comments to the -00 version. Mayutan Arumaithurai for his editing help with the -00 version. Poikselka Miikka, Isomaki Markus, Jari Mutikainen, Jean-Marie Stupka, and Antti Laurila for their comments and for pointing us to specifications outside the IETF. This document intentionally re-uses concept from existing documents. In particular, we reused ideas from SIEVE , a mail filtering language. the text in is based on the description in the Call Processing Language (CPL) . In general, the difference between CPL and this document is that CPL has a more procedural approach, while this proposal is matching-based. It is obviously possible to enhance CPL as well to provide the functionality offered in this document. text in from . content of , and is reused from .