Framework for Session Set-up with Media Authorization
How a call setup is linked to a resource reservation so that only authorized media gets bandwidth
Hamer, L-N., Gage, B., and H. Shieh, "Framework for Session Set-up with Media Authorization", RFC 3521, DOI 10.17487/RFC3521, April 2003, <https://www.rfc-editor.org/info/rfc3521>.
@misc{rfc3521,
series = {Request for Comments},
number = 3521,
howpublished = {RFC 3521},
publisher = {RFC Editor},
doi = {10.17487/RFC3521},
url = {https://www.rfc-editor.org/info/rfc3521},
author = {L-N. Hamer and B. Gage and H. Shieh},
title = {{Framework for Session Set-up with Media Authorization}},
pagetotal = 25,
year = 2003,
month = apr,
} RFC text
Contents
Network Working Group L-N. Hamer
Request for Comments: 3521 B. Gage
Category: Informational Nortel Networks
H. Shieh
AT&T Wireless
April 2003
Framework for Session Set-up with Media Authorization
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 Notice
Copyright (C) The Internet Society (2003). All Rights Reserved.
Abstract
Establishing multimedia streams must take into account requirements
for end-to-end QoS, authorization of network resource usage and
accurate accounting for resources used. During session set up,
policies may be enforced to ensure that the media streams being
requested lie within the bounds of the service profile established
for the requesting host. Similarly, when a host requests resources
to provide a certain QoS for a packet flow, policies may be enforced
to ensure that the required resources lie within the bounds of the
resource profile established for the requesting host.
To prevent fraud and to ensure accurate billing, this document
describes various scenarios and mechanisms that provide the linkage
required to verify that the resources being used to provide a
requested QoS are in-line with the media streams requested (and
authorized) for the session.
Table of Contents
1. Introduction....................................................2
2. Conventions used in this document...............................3
3. Definition of terms.............................................4
4. The Coupled Model...............................................5
4.1 Coupled Model Message Flows...............................6
4.2 Coupled Model Authorization Token.........................8
4.3 Coupled Model Protocol Impacts............................8
5. The Associated Model <<using One Policy Server>>................8
5.1 Associated Model Message Flows
<<using One Policy Server>>...............................9
5.2 Associated Model Authorization Token
<<using One Policy Server>>..............................11
5.3 Associated Model Protocol Impacts
<<using One Policy Server>>..............................11
5.4 Associated Model Network Impacts
<<using One Policy Server>>..............................12
6. The Associated Model <<using Two Policy Servers>>..............12
6.1 Associated Model Message Flows
<<using Two Policy Servers>>.............................13
6.2 Associated Model Authorization Token
<<using Two Policy Servers>>.............................15
6.3 Associated Model Protocol Impacts
<<using Two Policy Servers>>.............................16
7. The Non-Associated Model........................................16
7.1 Non-Associated Model Message Flow........................17
7.2 Non-Associated Model Authorization Token.................19
7.3 Non-Associated Model Protocol Impacts....................19
8. Conclusions....................................................20
9. Security Considerations........................................21
10. Normative References...........................................22
11. Informative References.........................................23
12. Acknowledgments................................................23
13. Authors' Addresses.............................................24
14. Full Copyright Statement.......................................25
1. Introduction
Various mechanisms have been defined through which end hosts can use
a session management protocol (e.g., SIP [6]) to indicate that QoS
requirements must be met in order to successfully set up a session.
However, a separate protocol (e.g., RSVP [7]) is used to request the
resources required to meet the end-to-end QoS of the media stream.
To prevent fraud and to ensure accurate billing, some linkage is
required to verify that the resources being used to provide the
requested QoS are in-line with the media streams requested (and
authorized) for the session.
This document describes such a linkage through use of a "token" that
provides capabilities similar to that of a gate in [12] and of a
ticket in the push model of [10]. The token is generated by a policy
server (or a session management server) and is transparently relayed
through the end host to the edge router where it is used as part of
the policy-controlled flow admission process.
In some environments, authorization of media streams can exploit the
fact that pre-established relationships exist between elements of the
network (e.g., session management servers, edge routers, policy
servers and end hosts). Pre-established relationships assume that
the different network elements are configured with the identities of
the other network elements and, if necessary, are configured with
security keys, etc. required to establish a trust relationship. In
other environments, however, such pre-established relationships may
not exist either due to the complexity of creating these associations
a priori (e.g., in a network with many elements), or due to the
different business entities involved (e.g., service provider and
access provider), or due to the dynamic nature of these associations
(e.g., in a mobile environment).
In this document, we describe these various scenarios and the
mechanisms used for exchanging information between network elements
in order to authorize the use of resources for a service and to
coordinate actions between the session and resource management
entities. Specific extensions to session management protocols (e.g.,
SIP [6], H.323), to resource reservation protocols (e.g., RSVP [4],
YESSIR) and to policy management protocols (e.g., COPS-PR [9], COPS-
RSVP [3]) required to realize these scenarios and mechanisms are
beyond the scope of this document.
For clarity, this document will illustrate the media authorization
concepts using SIP for session signalling, RSVP for resource
reservation and COPS for interaction with the policy servers. Note,
however, that the framework could be applied to a multimedia services
scenario using different signalling protocols.
2. Conventions used in this document
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 BCP 14, RFC 2119 [1].
3. Definition of terms
Figure 1 introduces a generic model for session establishment, QoS
and policy enforcement.
+-------------------------------------+ +---+
| SCD - Service Control Domain | | |
| +-----------------------+ +--------+| | I |
| |Session management | |Policy || | n |
| |server | |Server || | t |
| | +---------+ +------+ | | +----+||<->| e |
| | |SIP Proxy| |PEP |<-|-|->|PDP ||| | r |
| | +---------+ +------+ | | +----+|| | - |
| +-----------------------+ +--------+| | c |
| | | o |
+-------------------------------------+ | n |
| n |
+-------------------------------------+ | e |
| RCD - Resource Control Domain | | c |
| | | t |
| | | i |
| +------------+ +-------------+ | | n |
+----------+ | |Edge Router | |Policy Server| | | g |
| End | | | | | | | | |
| Host | | |+----------+| |+----------+ | | | N |
|+--------+| | ||RSVP Agent|| ||PDP | | | | e |
||RSVP ||<->| |+----------+|<-->|+----------+ | |<->| t |
||Client || | |+----------+| | | | | w |
|+--------+| | || PEP || | | | | o |
||SIP User|| | |+----------+| | | | | r |
||Agent || | +------------+ +-------------+ | | k |
|+--------+| | | | |
+----------+ +-------------------------------------+ +---+
Figure 1: Generic media authorization network model
EH - End Host: The End Host is a device used by a subscriber to
access network services. The End Host includes a client for
requesting network services (e.g., through SIP) and a client for
requesting network resources (e.g., through RSVP).
ER - Edge Router: The Edge Router is a network element connecting the
end host to the rest of the Resource Control Domain. The Edge Router
contains a PEP to enforce policies related to resource usage in the
Resource Control Domain by the End Host. It also contains a
signalling agent (e.g., for RSVP) for handling resource reservation
requests from the End Host.
PDP - Policy Decision Point: The PDP is a logical entity located in
the Policy Server that is responsible for authorizing or denying
access to services and/or resources.
PEP - Policy Enforcement Point: The PEP is a logical entity that
enforces policy decisions made by the PDP. Note that other PEPs may
reside in other network elements not shown in the model of Figure 1,
however they will not be discussed in this document.
PS - Policy Server: The Policy Server is a network element that
includes a PDP. Note that there may be a PS in the Service Control
Domain to control use of services and there may be a separate PS in
the Resource Control Domain to control use of resources along the
packet forwarding path. Note also that network topology may require
multiple Policy Servers within either Domain, however they provide
consistent policy decisions to offer the appearance of a single PDP
in each Domain.
RCD - Resource Control Domain: The Resource Control Domain is a
logical grouping of elements that provide connectivity along the
packet forwarding paths to and from an End Host. The RCD contains ER
and PS entities whose responsibilities include management of
resources along the packet forwarding paths. Note that there may be
one or more RCDs within an autonomous domain.
SCD - Service Control Domain: The Service Control Domain is a logical
grouping of elements that offer applications and content to
subscribers of their services. The Session Management Server resides
in the SCD along with a PS. Note that there may be one or more SCDs
within an autonomous domain.
SMS - Session Management Server: The Session Management Server is a
network element providing session management services (e.g.,
telephony call control). The Session Management Server contains a
PEP to enforce policies related to use of services by the End Host.
It also contains a signalling agent or proxy (e.g., for SIP) for
handling service requests from the End Host.
4. The Coupled Model
In some environments, a pre-established trust relationship exists
between elements of the network (e.g., session management servers,
edge routers, policy servers and end hosts). We refer to this as the
"coupled model", indicating the tight relationship between entities
that is presumed. The key aspects of this scenario are the
following:
- Policy decisions, including media authorization, are made by a
single Policy Server.
- The Edge Router, Session Management Servers and Policy Server
involved in establishing the session are known a priori. For
example, the End Host may be configured to use a Session
Management Server associated with the Edge Router to which the EH
is connected.
- There are pre-defined trust relationships between the SMS and the
PS and between the ER and the PS.
+--------+
+------+ | |
| | 1 +--------------------+ 2 | |
| |-------->| Session Management |----->| |
| |<--------| Server |<-----| |
| | 4 +--------------------+ 3 | |
| End | | Policy |
| Host | | Server |
| | | |
| | 5 +--------------------+ 6 | |
| |-------->| Edge |----->| |
| |<--------| Router |<-----| |
| | 8 +--------------------+ 7 | |
+------+ | |
+--------+
Figure 2: The Coupled Model
4.1 Coupled Model Message Flows
In this model, it is assumed that there is one Policy Server serving
both the Service Control and Resource Control Domains and that there
are pre-defined trust relationships between the PS and SMS and
between the PS and ER. Communications between these entities are
then possible as described below. Only the originating side flows
are described for simplicity. The same concepts apply to the
terminating side.
1. The End Host issues a session set-up request (e.g., SIP INVITE) to
the Session Management Server indicating, among other things, the
media streams to be used in the session. As part of this step,
the End Host may authenticate itself to the Session Management
Server.
2. The Session Management Server, possibly after waiting for
negotiation of the media streams to be completed, sends a policy
decision request (e.g., COPS REQ) to the Policy Server in order to
determine if the session set-up request should be allowed to
proceed.
3. The Policy Server sends a decision (e.g., COPS DEC) to the Session
Management Server, possibly after modifying the parameters of the
media to be used. Included in this response is a "token" that can
subsequently be used by the Policy Server to identify the session
and the media it has authorized.
4. The Session Management Server sends a response to the End Host
(e.g., SIP 200 or 183) indicating that session set-up is complete
or is progressing. Included in this response is a description of
the negotiated media along with the token from the Policy Server.
5. The End Host issues a request (e.g., RSVP PATH) to reserve the
resources necessary to provide the required QoS for the media
stream. Included in this request is the token from the Policy
Server provided via the Session Management Server.
6. The Edge Router intercepts the reservation request and sends a
policy decision request (e.g., COPS REQ) to the Policy Server in
order to determine if the resource reservation request should be
allowed to proceed. Included in this request is the token from
the Policy Server provided by the End Host. The Policy Server
uses this token to correlate the request for resources with the
media authorization previously provided to the Session Management
Server.
7. The Policy Server sends a decision (e.g., COPS DEC) to the Edge
Router, possibly after modifying the parameters of the resources
to be reserved.
8. The Edge Router, possibly after waiting for end-to-end negotiation
for resources to be completed, sends a response to the End Host
(e.g., RSVP RESV) indicating that resource reservation is complete
or is progressing.
4.2 Coupled Model Authorization Token
In the Coupled Model, the Policy Server is the only network entity
that needs to interpret the contents of the token. Therefore, in
this model, the contents of the token are implementation dependent.
Since the End Host is assumed to be untrusted, the Policy Server
SHOULD take measures to ensure that the integrity of the token is
preserved in transit; the exact mechanisms to be used are also
implementation dependent.
4.3 Coupled Model Protocol Impacts
The use of a media authorization token in the Coupled Model requires
the addition of new fields to several protocols:
- Resource reservation protocol. A new protocol field or object
MUST be added to the resource reservation protocol to
transparently transport the token from the End Host to the Edge
Router. The content and internal structure (if any) of this
object SHOULD be opaque to the resource reservation protocol. For
example, this is achieved in RSVP with the Policy Data object
defined in [8].
- Policy management protocol. A new protocol field or object MUST
be added to the policy management protocol to transparently
transport the token from the Policy Server to the Session
Management Server and from the Edge Router to the Policy Server.
The content and internal structure (if any) of this object SHOULD
be opaque to the policy management protocol. For example, this is
achieved in COPS-RSVP with the Policy Data object defined in [8].
- Session management protocol. A new protocol field or object MUST
be added to the session management protocol to transparently
transport the media authorization token from the Session
Management Server to the End Host. The content and internal
structure (if any) of this object SHOULD be opaque to the session
management protocol (e.g., SIP [6]).
5. The Associated Model <<using One Policy Server>>
In this scenario, there are multiple instances of the Session
Management Servers, Edge Routers and Policy Servers. This leads to a
network of sufficient complexity that it precludes distributing
knowledge of network topology to all network entities. The key
aspects of this scenario are the following:
- Policy decisions, including media authorization, are made by the
same Policy Server for both the Session Management Server and the
Edge Router. However, the Policy Server may change on a per-
transaction basis, i.e., on a per policy request basis.
- The Edge Router, Session Management Server and Policy Server
involved in establishing the session are not known a priori. For
example, the End Host may be dynamically configured to use one of
a pool of Session Management Servers and each of the Session
Management Servers may be statically configured to use one of a
pool of Policy Servers.
In another example, the End Host may be mobile and continually
changing the Edge Router that its point of attachment uses to
communicate with the rest of the network.
- There are pre-defined trust relationships between the SMS and the
PS and between the ER and the PS.
+---------------------+ +---------+
| SMS 'n' |<-->| PS 'm' |
+---------------------+ +--------+ |
+------+ : : : | | |
| | 1 +--------------------+ 2 | | |
| |-------->| Session Management |----->| | |
| |<--------| Server 1 |<-----| | |
| | 4 +--------------------+ 3 | | |
| End | | Policy | |
| Host | +--------------------+ | Server | |
| | | ER 'n' | | 1 | |
| | 5 +-+------------------+ | | | |
| |-------->| Edge |-+ 6 | | |
| |<--------| Router |----->| | |
| | 8 +--------------------+ 7 | | |
+------+ <-----| |-+
+--------+
Figure 3: The Associated Model using One Policy Server
5.1 Associated Model Message Flows <<using One Policy Server>>
In this model, it is assumed that a Policy Server can make decisions
for both the Service Control and Resource Control Domains and that
there are pre-defined trust relationships between the PS and SMS and
between the PS and ER. Communications between these entities are
then possible as described below. Only the originating side flows
are described for simplicity. The same concepts apply to the
terminating side.
1. The End Host issues a session set-up request (e.g., SIP INVITE) to
the Session Management Server indicating, among other things, the
media streams to be used in the session. As part of this step,
the End Host may authenticate itself to the Session Management
Server.
2. The Session Management Server, possibly after waiting for
negotiation of the media streams to be completed, sends a policy
decision request (e.g., COPS REQ) to the Policy Server in order to
determine if the session set-up request should be allowed to
proceed.
3. The Policy Server sends a decision (e.g., COPS DEC) to the Session
Management Server, possibly after modifying the parameters of the
media to be used. Included in this response is a "token" that can
subsequently be used by the Policy Server to identify the session
and the media it has authorized.
4. The Session Management Server sends a response to the End Host
(e.g., SIP 200 or 183) indicating that session set-up is complete
or is progressing. Included in this response is a description of
the negotiated media along with the token from the Policy Server.
5. The End Host issues a request (e.g., RSVP PATH) to reserve the
resources necessary to provide the required QoS for the media
stream. Included in this request is the token from the Policy
Server provided via the Session Management Server.
6. The Edge Router intercepts the reservation request and inspects
the token to learn which Policy Server authorized the media. It
then sends a policy decision request to that Policy Server in
order to determine if the resource reservation request should be
allowed to proceed. Included in this request is the token from
the Policy Server provided by the End Host. The Policy Server
uses this token to correlate the request for resources with the
media authorization previously provided to the Session Management
Server.
7. The Policy Server sends a decision to the Edge Router, possibly
after modifying the parameters of the resources to be reserved.
8. The Edge Router, possibly after waiting for end-to-end negotiation
for resources to be completed, sends a response to the End Host
(e.g., RSVP RESV) indicating that resource reservation is complete
or is progressing.
5.2 Associated Model Authorization Token <<using One Policy Server>>
Since the ER does not know which SMS and PS are involved in session
establishment, the token MUST include:
- A correlation identifier. This is information that the Policy
Server can use to correlate the resource reservation request with
the media authorized during session set up. The Policy Server is
the only network entity that needs to interpret the contents of
the correlation identifier therefore, in this model, the contents
of the correlation identifier are implementation dependent. Since
the End Host is assumed to be untrusted, the Policy Server SHOULD
take measures to ensure that the integrity of the correlation
identifier is preserved in transit; the exact mechanisms to be
used are also implementation dependent.
- The identity of the authorizing entity. This information is used
by the Edge Router to determine which Policy Server should be used
to solicit resource policy decisions.
In some environments, an Edge Router may have no means for
determining if the identity refers to a legitimate Policy Server
within its domain. In order to protect against redirection of
authorization requests to a bogus authorizing entity, the token
SHOULD also include:
- Authentication data. This authentication data is calculated over
all other fields of the token using an agreed mechanism. The
mechanism used by the Edge Router is beyond the scope of this
document.
The detailed semantics of an authorization token are defined in [4].
5.3 Associated Model Protocol Impacts <<using One Policy Server>>
The use of a media authorization token in this version of the
Associated Model requires the addition of new fields to several
protocols:
- Resource reservation protocol. A new protocol field or object
MUST be added to the resource reservation protocol to
transparently transport the token from the End Host to the Edge
Router. The content and internal structure of this object MUST be
specified so that the Edge Router can distinguish between the
elements of the token described in Section 5.2. For example, this
is achieved in RSVP with the Policy Data object defined in [8].
- Policy management protocol. A new protocol field or object MUST
be added to the policy management protocol to transparently
transport the token -- or at least the correlation identifier --
from the Edge Router to the Policy Server. The content and
internal structure of this object SHOULD be opaque to the policy
management protocol. For example, this is achieved in COPS-RSVP
with the Policy Data object defined in [8].
- Session management protocol. A new protocol field or object MUST
be added to the session management protocol to transparently
transport the media authorization token from the Session
Management Server to the End Host. The content and internal
structure of this object SHOULD be opaque to the session
management protocol (e.g., SIP [6]).
5.4 Associated Model Network Impacts <<using One Policy Server>>
The use of a media authorization token in this version of the
Associated Model requires that the Edge Router inspect the token to
learn which Policy Server authorized the media. In some
environments, it may not be possible for the Edge Router to perform
this function; in these cases, an Associated Model using Two Policy
Servers (section 6) is required.
This version of the Associated Model also requires that the Edge
Router interact with multiple Policy Servers. Policy decisions are
made by the same Policy Server for both the Session Management Server
and the Edge Router, however the Policy Server may change on per-
transaction basis. Note that the COPS framework does not currently
allow PEPs to change PDP on a per-transaction basis. To use this
model, a new framework must be defined for policy decision
outsourcing. This model also implies that the Policy Servers are
able to interact and/or make decisions for the Edge Router in a
consistent manner (e.g., as though there is only a single RCD Policy
Server). How this is accomplished is beyond the scope of this
document.
6. The Associated Model <<using Two Policy Servers>>
In this scenario, there are multiple instances of the Session
Management Servers, Edge Routers and Policy Servers. This leads to a
network of sufficient complexity that it precludes distributing
knowledge of network topology to all network entities. The key
aspects of this scenario are the following:
- Policy decisions, including media authorization, are made by
Policy Servers.
- There is a PS in the Resource Control Domain that is separate from
the PS in the Service Control Domain.
- The Edge Router, Session Management Server and Policy Servers
involved in establishing the session are not known a priori. For
example, the End Host may be dynamically configured to use one of
a pool of Session Management Servers or the End Host may be mobile
and continually changing the Edge Router that it uses to
communicate with the rest of the network.
- There is a pre-defined trust relationship between the SMS and the
SCD PS.
- There is a pre-defined trust relationship between the ER and the
RCD PS.
- There is a pre-defined trust relationship between the RCD and SCD
Policy Servers.
+--------------------+ +--------+
+------+ | SMS `n' | | |
| | 1 +-+------------------+ | | SCD |
| |-------->| Session Management |-+ 2 | Policy |
| |<--------| Server |----->| Server |
| | 4 +--------------------+<-----| |
| End | 3 +--------+
| | 7 ^ |
| Host | +--------------------+ | v 8
| | | ER 'n' | +--------+
| | 5 +-+------------------+ | | |
| |-------->| Edge |-+ 6 | RCD |
| |<--------| Router |----->| Policy |
| | 10 +--------------------+<--- -| Server |
+------+ 9 | |
+--------+
Figure 4: The Associated Model using Two Policy Servers
6.1 Associated Model Message Flows <<using Two Policy Servers>>
In this model, it is assumed that there is one Policy Server for the
Service Control Domain and a different Policy Server for the Resource
Control Domain. There are pre-defined trust relationships between
the SCD PS and SMS, between the RCD PS and ER and between the RCD and
SCD Policy Servers. Communications between these entities are then
possible as described below. Only the originating side flows are
described for simplicity. The same concepts apply to the terminating
side.
1. The End Host issues a session set-up request (e.g., SIP INVITE)
to the Session Management Server indicating, among other things,
the media streams to be used in the session. As part of this
step, the End Host may authenticate itself to the Session
Management Server.
2. The Session Management Server, possibly after waiting for
negotiation of the media streams to be completed, sends a policy
decision request (e.g., COPS REQ) to the SCD Policy Server in
order to determine if the session set-up request should be
allowed to proceed.
3. The SCD Policy Server sends a decision (e.g., COPS DEC) to the
Session Management Server, possibly after modifying the
parameters of the media to be used. Included in this response is
a "token" that can subsequently be used by the SCD Policy Server
to identify the session and the media it has authorized.
4. The Session Management Server sends a response to the End Host
(e.g., SIP 200 or 183) indicating that session set-up is complete
or is progressing. Included in this response is a description of
the negotiated media along with the token from the SCD Policy
Server.
5. The End Host issues a request (e.g., RSVP PATH) to reserve the
resources necessary to provide the required QoS for the media
stream. Included in this request is the token from the SCD
Policy Server provided via the Session Management Server.
6. The Edge Router intercepts the reservation request and sends a
policy decision request (e.g., COPS REQ) to the RCD Policy Server
in order to determine if the resource reservation request should
be allowed to proceed. Included in this request is the token
from the SCD Policy Server provided by the End Host.
7. The RCD Policy Server uses this token to learn which SCD Policy
Server authorized the media. It then sends an authorization
request [11] to that SCD Policy Server in order to determine if
the resource reservation request should be allowed to proceed.
Included in this request is the token from the SCD Policy Server
provided by the End Host.
8. The SCD Policy Server uses this token to correlate the request
for resources with the media authorization previously provided to
the Session Management Server. The SCD Policy Server sends a
decision [11] to the RCD Policy Server on whether the requested
resources are within the bounds authorized by the SCD Policy
Server.
9. The RCD Policy Server sends a decision (e.g., COPS DEC) to the
Edge Router, possibly after modifying the parameters of the
resources to be reserved.
10. The Edge Router, possibly after waiting for end-to-end
negotiation for resources to be completed, sends a response to
the End Host (e.g., RSVP RESV) indicating that resource
reservation is complete or is progressing
6.2 Associated Model Authorization Token <<using Two Policy Servers>>
Since the RCD Policy Server does not know which SMS and SCD PS are
involved in session establishment, the token MUST include:
- A correlation identifier. This is information that the SCD Policy
Server can use to correlate the resource reservation request with
the media authorized during session set up. The SCD Policy Server
is the only network entity that needs to interpret the contents of
the correlation identifier therefore, in this model, the contents
of the correlation identifier are implementation dependent. Since
the End Host is assumed to be untrusted, the SCD Policy Server
SHOULD take measures to ensure that the integrity of the
correlation identifier is preserved in transit; the exact
mechanisms to be used are also implementation dependent.
- The identity of the authorizing entity. This information is used
by the RCD Policy Server to determine which SCD Policy Server
should be used to verify the contents of the resource reservation
request.
In some environments, an RCD Policy Server may have no means for
determining if the identity refers to a legitimate SCD Policy Server.
In order to protect against redirection of authorization requests to
a bogus authorizing entity, the token SHOULD include:
- Authentication data. This authentication data is calculated over
all other fields of the token using an agreed mechanism. The
mechanism used by the RCD Policy Server is beyond the scope of
this document.
Note that the information in this token is the same as that in
Section 5.2 for the "One Policy Server" scenario.
The detailed semantics of an authorization token are defined in [4].
6.3 Associated Model Protocol Impacts <<using Two Policy Servers>>
The use of a media authorization token in this version of the
Associated Model requires the addition of new fields to several
protocols:
- Resource reservation protocol. A new protocol field or object
MUST be added to the resource reservation protocol to
transparently transport the token from the End Host to the Edge
Router. The content and internal structure of this object SHOULD
be opaque to the resource reservation protocol. For example, this
is achieved in RSVP with the Policy Data object defined in [8].
- Policy management protocol. A new protocol field or object MUST
be added to the policy management protocol to transport the token
from the SCD Policy Server to the Session Management Server and
from the Edge Router to the RCD Policy Server. The content and
internal structure of this object MUST be specified so that the
Policy Servers can distinguish between the elements of the token
described in Section 6.2. For example, this is achieved in COPS-
RSVP with the Policy Data object defined in [8].
- Session management protocol. A new protocol field or object MUST
be added to the session management protocol to transparently
transport the media authorization token from the Session
Management Server to the End Host. The content and internal
structure of this object SHOULD be opaque to the session
management protocol (e.g., SIP [6]).
Note that these impacts are the same as those discussed in Section
5.3 for the "One Policy Server" scenario. However the use of two
Policy Servers has one additional impact:
- Authorization protocol. A new protocol field or object MUST be
added to the authorization protocol to transport the token from
the RCD Policy Server to the SCD Policy Server. The content and
internal structure of this object MUST be specified so that the
Policy Servers can distinguish between the elements of the token
described in Section 6.2.
7. The Non-Associated Model
In this scenario, the Session Management Servers and Edge Routers are
associated with different Policy Servers, the network entities do not
have a priori knowledge of the topology of the network and there are
no pre-established trust relationships between entities in the
Resource Control Domain and entities in the Service Control Domain.
The key aspects of this scenario are the following:
- Policy decisions, including media authorization, are made by
Policy Servers.
- The PS in the Resource Control Domain is separate from the PS in
the Service Control Domain.
- There is a pre-defined trust relationship between the SMS and the
SCD PS.
- There is a pre-defined trust relationship between the ER and the
RCD PS.
- There are no pre-defined trust relationships between the ER and
SMS or between the RCD and SCD Policy Servers.
+--------+
+------+ | |
| | 1 +--------------------+ 2 | SCD |
| |-------->| Session Management |----->| Policy |
| |<--------| Server |<-----| Server |
| | 4 +--------------------+ 3 | |
| End | +--------+
| Host |
| | +--------+
| | 5 +--------------------+ 6 | |
| |-------->| Edge |----->| RCD |
| |<--------| Router |<-----| Policy |
| | 8 +--------------------+ 7 | Server |
+------+ | |
+--------+
Figure 5: The Non-Associated Model
7.1 Non-Associated Model Message Flow
In this model it is assumed that the policy servers make independent
decisions for their respective domains, obviating the need for
information exchange between policy servers. This model also enables
session authorization when communication between policy servers is
not possible for various reasons. It may also be used as a means to
speed up session setup and still ensure proper authorization is
performed.
This model does not preclude the possibility that the policy servers
may communicate at other times for other purposes (e.g., exchange of
accounting information).
Communications between network entities in this model is described
below. Only the originating side flows are described for simplicity.
The same concepts apply to the terminating side.
1. The End Host issues a session set-up request (e.g., SIP INVITE) to
the Session Management Server indicating, among other things, the
media streams to be used in the session. As part of this step,
the End Host may authenticate itself to the Session Management
Server.
2. The Session Management Server, possibly after waiting for
negotiation of the media streams to be completed, sends a policy
decision request (e.g., COPS REQ) to the SCD Policy Server in
order to determine if the session set-up request should be allowed
to proceed.
3. The SCD Policy Server sends a decision (e.g., COPS DEC) to the
Session Management Server, possibly after modifying the parameters
of the media to be used. Included in this response is a "token"
that can subsequently be used by the RCD Policy Server to
determine what media has been authorized.
4. The Session Management Server sends a response to the End Host
(e.g., SIP 200 or 183) indicating that session set-up is complete
or is progressing. Included in this response is a description of
the negotiated media along with the token from the SCD Policy
Server.
5. The End Host issues a request (e.g., RSVP PATH) to reserve the
resources necessary to provide the required QoS for the media
stream. Included in this request is the token from the SCD Policy
Server provided via the Session Management Server.
6. The Edge Router intercepts the reservation request and sends a
policy decision request (e.g., COPS REQ) to the RCD Policy Server
in order to determine if the resource reservation request should
be allowed to proceed. Included in this request is the token from
the SCD Policy Server provided by the End Host.
7. The RCD Policy Server uses this token to extract information about
the media that was authorized by the SCD Policy Server. The RCD
Policy Server uses this information in making its decision on
whether the resource reservation should be allowed to proceed.
The Policy Server sends a decision (e.g., COPS DEC) to the Edge
Router, possibly after modifying the parameters of the resources
to be reserved.
8. The Edge Router, possibly after waiting for end-to-end negotiation
for resources to be completed, sends a response to the End Host
(e.g., RSVP RESV) indicating that resource reservation is complete
or is progressing
7.2 Non-Associated Model Authorization Token
In this model, the token MUST contain sufficient information to allow
the RCD Policy Server to make resource policy decisions autonomously
from the SCD Policy Server. The token is created using information
about the session received by the SMS. The information in the token
MUST include:
- Calling party name or IP address (e.g., from SDP "c=" parameter).
- Called party name or IP address (e.g., from SDP "c=" parameter).
- The characteristics of (each of) the media stream(s) authorized
for this session (e.g., codecs, maximum bandwidth from SDP "m="
and/or "b=" parameters).
- The authorization lifetime. To protect against replay attacks,
the token should be valid for only a few seconds after the start
time of the session.
- The identity of the authorizing entity to allow for validation of
the token.
- Authentication data used to prevent tampering with the token.
This authentication data is calculated over all other fields of
the token using an agreed mechanism. The mechanism used by the
RCD Policy Server is beyond the scope of this document.
Furthermore, the token MAY include:
- The lifetime of (each of) the media stream(s) (e.g., from SDP "t="
parameter). This field may be useful in pre-paid scenarios in
order to limit the lifetime of the session.
- The Calling and called party port numbers (e.g., from the "m="
parameter).
The detailed semantics of an authorization token are defined in [4].
7.3 Non-Associated Model Protocol Impacts
The use of a media authorization token in the Non-Associated Model
requires the addition of new fields to several protocols:
- Resource reservation protocol. A new protocol field or object
MUST be added to the resource reservation protocol to
transparently transport the token from the End Host to the Edge
Router. The content and internal structure of this object SHOULD
be opaque to the resource reservation protocol. For example, this
is achieved in RSVP with the Policy Data object defined in [8].
- Policy management protocol. A new protocol field or object MUST
be added to the policy management protocol to transport the token
from the SCD Policy Server to the Session Management Server and
from the Edge Router to the RCD Policy Server. The content and
internal structure of this object MUST be specified so that the
Policy Servers can distinguish between the elements of the token
described in Section 7.2. For example, this is achieved in COPS-
RSVP with the Policy Data object defined in [8].
- Session management protocol. A new protocol field or object MUST
be added to the session management protocol to transparently
transport the media authorization token from the Session
Management Server to the End Host. The content and internal
structure of this object SHOULD be opaque to the session
management protocol (e.g., SIP [6]).
8. Conclusions
This document defines three models for session set-up with media
authorization:
- The Coupled Model which assumes a priori knowledge of network
topology and where pre-established trust relationships exist
between network entities.
- The Associated Model where there are common or trusted policy
servers but knowledge of the network topology is not known a
priori.
- The Non-Associated Model where knowledge of the network topology
is not known a priori, where there are different policy servers
involved and where a trust relationship does not exist between the
policy servers.
The Associated Model is applicable to environments where the network
elements involved in establishing a session have a pre-determined
trust relationship but where their identities must be determined
dynamically during session set up. The Non-Associated Model is
applicable to environments where there is a complex network topology
and/or where trust relationships between domains do not exist (e.g.,
when they are different business entities).
In any given network, one or more of these models may be applicable.
Indeed, the model to be used may be chosen dynamically during session
establishment based on knowledge of the end points involved in the
call. In all cases, however, there is no need for the End Host or
the Session Management Server to understand or interpret the
authorization token - to them it is an opaque protocol element that
is simply copied from one container protocol to another.
Finally, the framework defined in this document is extensible to any
kind of session management protocol coupled to any one of a number of
resource reservation and/or policy management protocols.
9. Security Considerations
The purpose of this document is to describe a mechanism for media
authorization to prevent theft of service.
For the authorization token to be effective, its integrity MUST be
guaranteed as it passes through untrusted network entities such as
the End Host. This can be achieved by using authentication data.
There is no requirement for encryption of the token since it does not
contain confidential information that may be used by malicious users.
This document assumes that trust relationships exist between various
network entities, as described in each of the models. The means for
establishing these relationships are beyond the scope of this
document.
The different interfaces between the network entities described in
this document have different natures requiring different security
characteristics:
- The edge router and RCD policy server MUST have a trust
relationship. If necessary, this relationship can be enforced
through a formal security association [14].
- The network policies exchanged over the interface between edge
router and RCD policy server SHOULD be integrity protected. This
can be accomplished using integrity mechanisms built into the
policy control protocol (e.g., the Integrity object in COPS [2])
or through generic IP security mechanisms [14].
- The SCD and RCD policy servers MUST have a trust relationship in
the associated model. If necessary, this relationship can be
enforced through a formal security association [14].
- The information exchanged over the interface between policy
servers SHOULD be integrity protected. This can be accomplished
using integrity mechanisms built into the policy exchange protocol
[2] or through generic IP security mechanisms [14].
- The end host SHOULD be authenticated by the RCD to protect against
identity theft. The network resource request/responses should be
protected against corruption and spoofing. Thus, the interface
between host and edge router SHOULD provide integrity and
authentication of messages. For example, [13] provides integrity
and authentication of RSVP messages.
- The end host SHOULD be authenticated by the SCD to protect against
identity theft. The session setup request/response should be
protected against corruption and spoofing. Thus, the interface
between host and SMS SHOULD provide integrity and authentication
of messages.
- The SMS and the SCD policy server MUST have a a trust
relationship. If necessary, this relationship can be enforced
through a formal security association [14].
- The network policies exchanged over the interface between the SMS
and SCD policy server SHOULD be integrity protected. This can be
accomplished using integrity mechanisms built into the policy
control protocol (e.g., the Integrity object in COPS [2]) or
through generic IP security mechanisms [14].
10. Normative References
[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.
[2] Durham, D., Boyle, J., Cohen, R., Herzog, S., Rajan, R. and A.
Sastry, "The COPS (Common Open Policy Service) Protocol", RFC
2748, January 2000.
[3] Herzog, S., Boyle, J., Cohen, R., Durham, D., Rajan, R. and A.
Sastry, "COPS usage for RSVP", RFC 2749, January 2000.
[4] Hamer, L-N., Gage, B., Kosinski, B. and H. Shieh, "Session
Authorization Policy Element", RFC 3520, April 2003.
[5] Handley, M. and V. Jacobson, "SDP: session description
protocol," RFC 2327, April 1998.
[6] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A.,
Peterson, J., Sparks, R., Handley, M. and E. Schooler, "SIP:
Session Initiation Protocol", RFC 3261, June 2002.
[7] Braden, R., Zhang, L., Berson, S., Herzog, S. and S. Jamin,
"Resource ReSerVation protocol (RSVP) -- version 1 functional
specification," RFC 2205, September 1997.
[8] Herzog, S., "RSVP Extensions for Policy Control", RFC 2750,
January 2000.
[9] Chan, K., Seligson, J., Durham, D., Gai, S., McCloghrie, K.,
Herzog, S., Reichmeyer, F., Yavatkar, R. and A. Smith, "COPS
Usage for Policy Provisioning (COPS-PR)", RFC 3084, March 2001.
11. Informative References
[10] Vollbrecht, J., Calhoun, P., Farrell, S., Gommans, L., Gross,
G., de Bruijn, B., de Laat, C., Holdrege, M. and P. Spence, "AAA
Authorization Framework", RFC 2904, August 2000.
[11] de Laat, C., Gross, G., Gommans, L., Vollbrecht, J. and D.
Spence, "Generic AAA Architecture", RFC 2903, August 2000.
[12] "PacketCable Dynamic Quality of Service Specification",
CableLabs, December 1999.
[13] Baker, F., Lindell, B. and M. Talwar, "RSVP Cryptographic
Authentication", RFC 2747, January 2000.
[14] Kent, S. and R. Atkinson, "Security Architecture for the
Internet Protocol", RFC 2401, November 1998.
12. Acknowledgments
The authors would like to thank to following people for their useful
comments and suggestions related to this document: Kwok Ho Chan, Doug
Reeves, Sam Christie, Matt Broda, Yajun Liu, Brett Kosinski, Francois
Audet, Bill Marshall, Diana Rawlins and many others.
13. Authors' Addresses
Louis-Nicolas Hamer
Nortel Networks
PO Box 3511 Station C
Ottawa, ON
CANADA K1Y 4H7
Phone: +1 613.768.3409
EMail: nhamer@nortelnetworks.com
Bill Gage
Nortel Networks
PO Box 3511 Station C
Ottawa, ON
CANADA K1Y 4H7
Phone: +1 613.763.4400
EMail: gageb@nortelnetworks.com
Hugh Shieh
AT&T Wireless
7277 164th Avenue NE
Redmond, WA
USA 98073-9761
Phone: +1 425.580.6898
EMail: hugh.shieh@attws.com
14. Full Copyright Statement
Copyright (C) The Internet Society (2003). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS 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.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
ISO Plain Language
On this page
§1Why Session Setup Needs Authorization
Various mechanisms let an end host use a session management protocol such as SIP to indicate that quality of service requirements must be met to set up a session. A separate protocol, such as RSVP, is used to request the resources needed to meet the end-to-end quality of service of the media stream. To prevent fraud and to ensure accurate billing, some linkage is required to verify that the resources being used to provide the requested quality are in line with the media streams requested and authorized for the session.
The document describes that linkage through a token that gives capabilities similar to a gate or to a ticket. The token is generated by a policy server or a session management server, and is transparently relayed through the end host to the edge router, where it is used as part of the policy-controlled flow admission process. In some environments the authorization can rely on pre-established relationships between network elements, such as session management servers, edge routers, policy servers and end hosts that already know each other's identities and hold the keys needed for a trust relationship. In others such relationships may not exist, because creating them in advance is impractical in a network with many elements, because different business entities are involved, such as a service provider and an access provider, or because the associations change, as in a mobile environment.
The specific protocol extensions needed to realise these scenarios are beyond the scope of the document. For clarity it illustrates the concepts with SIP for session signalling, RSVP for resource reservation and COPS for interaction with the policy servers, while noting that the framework could be applied with other signalling protocols.
The call setup and the bandwidth reservation are two separate conversations, so something has to link them. This document uses a token: a short piece of information from the policy server that says which media were approved, and that travels with the reservation request so the network can check it.
| Model | Policy servers | Topology known in advance | Trust between the domains |
|---|---|---|---|
| Coupled | One, for both domains | Yes | Pre-established trust relationships exist between network entities |
| Associated | One or two, shared or trusted | No | Trust relationships exist within each domain |
| Non-Associated | Separate in each domain | No | No trust relationship between the policy servers |
§3The Elements of the Model
The document defines the elements that take part. The End Host is the device a subscriber uses to reach network services; it includes a client for requesting network services, such as SIP, and a client for requesting network resources, such as RSVP. The Edge Router connects the end host to the rest of the Resource Control Domain and contains a Policy Enforcement Point that enforces the policies related to resource usage, as well as a signalling agent such as an RSVP agent for handling reservation requests from the end host.
The Policy Decision Point is the logical entity in the Policy Server responsible for authorizing or denying access to services and resources, and the Policy Enforcement Point is the logical entity that enforces the decisions the Policy Decision Point makes. The Policy Server is the network element that contains the Policy Decision Point; there may be one in the Service Control Domain controlling the use of services, and a separate one in the Resource Control Domain controlling the use of resources along the packet forwarding path.
The Resource Control Domain is the logical grouping of elements that provide connectivity along the packet forwarding paths to and from an end host, and the Service Control Domain is the logical grouping of elements that offer applications and content to subscribers. The Session Management Server sits in the Service Control Domain and provides session management services such as telephony call control; it contains a Policy Enforcement Point for the use of services and a signalling agent or proxy, such as a SIP proxy, for handling service requests from the end host.
The document names the parts: the subscriber's own device, the router at the edge of the network, a policy server that says yes or no, and the two domains the elements belong to, one for services and one for the network resources the service uses.
§4The Coupled Model
The coupled model applies where a pre-established trust relationship exists between the elements of the network. Policy decisions, including media authorization, are made by a single Policy Server. The Edge Router, the Session Management Servers and the Policy Server involved in establishing the session are known in advance, for example because the end host is configured to use a Session Management Server associated with the Edge Router it is connected to. There are pre-defined trust relationships between the Session Management Server and the Policy Server, and between the Edge Router and the Policy Server.
The message flow has eight steps. The end host issues a session set-up request, such as a SIP INVITE, to the Session Management Server, giving the media streams to be used and possibly authenticating itself. The Session Management Server, possibly after the media negotiation has completed, sends a policy decision request to the Policy Server to find out whether the set-up should proceed. The Policy Server sends a decision back, possibly after modifying the parameters of the media, and includes in it a token that the Policy Server can later use to identify the session and the media it authorized. The Session Management Server then answers the end host with a SIP 200 or 183, carrying the negotiated media and the token.
The end host next issues an RSVP PATH to reserve the resources for the media stream, including the token it received. The Edge Router intercepts that request and sends a policy decision request to the Policy Server, including the token, so that the Policy Server can correlate the resource request with the media authorization it gave earlier. The Policy Server sends a decision back, possibly after modifying the parameters of the resources to be reserved, and the Edge Router answers the end host with an RSVP RESV once the reservation is complete or progressing.
Because only the Policy Server interprets the token, its contents are implementation dependent in this model. Since the end host is assumed to be untrusted, the Policy Server should take measures to preserve the integrity of the token in transit. Three protocols need a new field or object to carry the token: the resource reservation protocol, the policy management protocol and the session management protocol.
In the simplest case everything about the network is known in advance and the policy server is the same one for both domains. The token then only has to mean something to that one server, so its format is left to the implementation.
- INVITEThe end host asks for a session and names the media streams.
- COPS REQThe server asks whether the session may proceed.
- COPS DEC with tokenThe policy server decides and returns a token naming the authorized media.
- 200 or 183 with tokenThe end host receives the negotiated media and the token.
- RSVP PATH with tokenThe end host reserves resources and passes the token on.
- COPS REQ with tokenThe edge router asks the policy server to check the reservation.
- COPS DECThe policy server correlates the request with the earlier authorization.
- RSVP RESVThe reservation completes.
§5The Associated Model
The associated model applies when there are several Session Management Servers, Edge Routers and Policy Servers, so that the network is too complex for every element to know the topology. The Edge Router, the Session Management Server and the Policy Server involved are not known in advance: the end host may be configured to use one of a pool of Session Management Servers, each of which may be configured to use one of a pool of Policy Servers, and the end host may be mobile and continually changing the Edge Router it attaches to. Trust relationships still exist between the Session Management Server and the Policy Server and between the Edge Router and the Policy Server.
With one Policy Server, the flow is the same as in the coupled model, with one difference: the Edge Router inspects the token to learn which Policy Server authorized the media, and sends its policy decision request to that server. The token MUST therefore include a correlation identifier, which the Policy Server uses to match the resource request to the media it authorized during session set-up, and the identity of the authorizing entity, which tells the Edge Router which Policy Server to ask. Where an Edge Router has no way of telling whether the identity refers to a legitimate Policy Server in its domain, the token SHOULD also include authentication data calculated over all the other fields of the token.
With two Policy Servers, one in the Service Control Domain and one in the Resource Control Domain, the trust relationships are between the Session Management Server and the Service Control Domain server, between the Edge Router and the Resource Control Domain server, and between the two Policy Servers. The Resource Control Domain server then makes the resource decision, using information the Service Control Domain server passed to it.
The document notes two consequences. The Edge Router has to be able to read the token, and in some environments it cannot, in which case the two-policy-server associated model is required. And the Edge Router has to work with several Policy Servers, with the one that handles a given transaction changing from request to request, while the COPS framework does not currently allow enforcement points to change their decision point on a per-transaction basis, so a new framework for policy decision outsourcing would be needed.
When the network is too big for every router to know the topology, the token has to carry more. It must say which policy server gave the approval and must let that server match the reservation to the approval later.
§7The Non-Associated Model
The non-associated model applies where the Session Management Servers and Edge Routers are associated with different Policy Servers, the elements do not know the topology in advance, and there are no pre-established trust relationships between entities in the Resource Control Domain and entities in the Service Control Domain. The Session Management Server and the Service Control Domain Policy Server trust each other, and the Edge Router and the Resource Control Domain Policy Server trust each other, but the Edge Router and the Session Management Server do not trust each other, and neither do the two Policy Servers.
Here the policy servers make independent decisions for their own domains, which removes the need for information exchange between them and allows session authorization when the two servers cannot talk to each other. It can also be used to speed up session setup while still authorizing properly.
Because the Resource Control Domain server has to make its resource decision on its own, the token MUST carry enough information for it to do so. It must contain the calling party name or IP address, the called party name or IP address, the characteristics of each authorized media stream such as the codecs and the maximum bandwidth, the authorization lifetime, which should be valid for only a few seconds after the start time of the session to protect against replay attacks, the identity of the authorizing entity so the token can be validated, and authentication data calculated over all the other fields to prevent tampering. It MAY also carry the lifetime of each media stream, which is useful in pre-paid scenarios, and the calling and called party port numbers.
The document's conclusion is that one or more of the three models may apply in any given network, and that the model may even be chosen during session establishment based on what is known about the end points. In every case the end host and the Session Management Server do not need to understand the token: to them it is an opaque element copied from one container protocol to another.
In the hardest case the two policy servers belong to different organisations and do not trust each other. The token then has to describe the whole authorized session, including the parties, the media and a short lifetime, so that the second server can decide on its own.
- Called and calling partyThe name or IP address of each party, for example from the SDP c= parameter.
- Media characteristicsThe codecs and the maximum bandwidth of each authorized media stream.
- Authorization lifetimeThe token should be valid for only a few seconds after the session start time, to protect against replay.
- Authorizing entityThe identity of the authorizing entity, so the token can be validated.
- Authentication dataData calculated over all the other fields of the token to prevent tampering.
§9Security Considerations
The purpose of the mechanism is to prevent theft of service. For the token to be effective, its integrity MUST be guaranteed as it passes through untrusted network entities such as the end host, which can be achieved by using authentication data. There is no requirement for encryption of the token, since it does not contain confidential information that could be used by malicious users.
The document assumes that trust relationships exist between the network entities in each model, and that the means for establishing them are beyond its scope. The interfaces between the entities need different security characteristics. The edge router and the Resource Control Domain policy server MUST have a trust relationship, enforceable through a formal security association, and the network policies exchanged over that interface SHOULD be integrity protected, using the integrity mechanisms built into the policy control protocol or generic IP security mechanisms. The Service Control Domain and Resource Control Domain policy servers MUST have a trust relationship in the associated model, and the information exchanged between them SHOULD be integrity protected. The Session Management Server and the Service Control Domain policy server MUST have a trust relationship, and the policies exchanged over that interface SHOULD be integrity protected.
The end host SHOULD be authenticated by both domains to protect against identity theft, and the resource request and response and the session setup request and response SHOULD be protected against corruption and spoofing, so the interfaces between the host and the edge router, and between the host and the Session Management Server, should provide integrity and authentication of messages.
The token must not be altered on its way through the network, which is what stops a subscriber from changing what was approved. It does not have to be secret, and the rest of the document's advice is about the trust relationships between the servers that handle it.