Multidimensional Security Policy Management and Enhancements for IP Security Policy

Versions: 00                                                            
Internet Draft                                        M. Condell, BBN
Expires May, 2002                                   November 14, 2001

        Multidimensional Security Policy Management and
            Enhancements for IP Security Policy

Status of this Memo

This document is an Internet-Draft and is in full conformance
with all provisions of Section 10 of RFC2026.

Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups.  Note that
other groups may also distribute working documents as

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."

The list of current Internet-Drafts can be accessed at

The list of Internet-Draft Shadow Directories can be accessed at

 Copyright Notice

   Copyright (C) The Internet Society (2001).  All Rights Reserved.


This document describes the requirements and architecture for supporting
security policy resolution among a coalition of partners.  It then
discusses how solutions necessary for the coalition resolution problem
may be utilized to improve the on-going development of an IPSP

Condell                                                         [page 1]

Internet Draft                MSME and IPSP                November 2001

Table of Contents

1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  2
  1.1 Definitions. . . . . . . . . . . . . . . . . . . . . . . . .  3

2.  MSME . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  3
  2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . .  3
  2.2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . .  4
  2.3 Requirements . . . . . . . . . . . . . . . . . . . . . . . .  5
  2.4 Architecture . . . . . . . . . . . . . . . . . . . . . . . .  6
    2.4.1 MSME Components. . . . . . . . . . . . . . . . . . . . .  7 Policy Level Agreement . . . . . . . . . . . . . . .  7 Policy Compilation . . . . . . . . . . . . . . . . .  7 Resolution . . . . . . . . . . . . . . . . . . . . .  8 Exchange Protocol. . . . . . . . . . . . . . . . . .  8 Reconciliation . . . . . . . . . . . . . . . . . . .  9 Monitoring . . . . . . . . . . . . . . . . . . . . .  10
    2.4.2 Partner-Dependent Components . . . . . . . . . . . . . .  10 Policy Management Tool . . . . . . . . . . . . . . .  10 Policy Language. . . . . . . . . . . . . . . . . . .  10 Local Policy Repositories. . . . . . . . . . . . . .  11 Policy Enforcement Points. . . . . . . . . . . . . .  11
    2.4.3 Data Flow. . . . . . . . . . . . . . . . . . . . . . . .  12

3.  Ideas for IPSP . . . . . . . . . . . . . . . . . . . . . . . .  13
  3.1 Late Name Binding. . . . . . . . . . . . . . . . . . . . . .  13
  3.2 Generalization . . . . . . . . . . . . . . . . . . . . . . .  15

4. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . .  16

5. References. . . . . . . . . . . . . . . . . . . . . . . . . . .  16

1.  Introduction

This document describes the requirements and architecture for the
Multidimensional Security Management and Enforcement (MSME) system.
MSME provides a means for members of a dynamic coalition with limited
trust between them to negotiate mutually agreeable policies that
enable mission relevant communications.  MSME's solution allows each
member to maintain its own internal policy requirements, policy
languages, and policy management systems while enabling them to
exchange and resolve policies with other members of the coalition.

While it may be early for the IPSP working group to design solutions
beyond IPsec policy resolution between two hosts and any intervening
gateways, there are several benefits to looking at the requirements
and potential solutions for more complex security policy management
environments.  It exposes issues with, and presents possible solutions
to, generalizing the IPSP work to security protocols other than IPsec.
It may also help the working group to design a solution that is more
extensible to other security policy management needs.

Condell                                                         [page 2]

Internet Draft                MSME and IPSP                November 2001

The next section describes the MSME system motivations, requirements,
and architecture.  It is followed by a discussion of some of the ideas
that may be incorporated into the IPSP work.

1.1 Definitions

The following terms are used throughout this document, in addition to
the terms defined for general policy terminology [TERM].


   A coalition is a group of administrative entities (e.g. companies,
   countries) that work together to achieve a defined objective
   (mission).  The coalition will have specific communications
   requirements necessary to accomplish the objective.


   A partner is an administrative entity that participates in a

"MAY" that appear in this document are to be interpreted as described
in [Bra97].

2   MSME

The MSME system is being developed under DARPA contract [Contract
Number F30602-00-C-0062].

2.1 Motivation

Policy management, while fundamental for good network security, is a
complex task.  It is difficult in the IPSP world where two end-hosts
are trying to communicate, but is even more complex when groups are
trying to communicate.  Coalitions are a particularly difficult
environment for policy management.  There may be a large number of
partners that need to share policy information.  Partners may join or
leave the coalition, each change possibly affecting the policy that
must be enforced.  Each partner may have different management elements,
including different policy languages, policy distribution protocols,
and policy authoring tools.  Each partner may even implement a
different set of security protocols.

Despite the heterogeneity of the coalition environment, it is
necessary to be able to determine if the coalition can implement the
security policies necessary to carry out its mission.  This requires
that the coalition partners be able to exchange their policies,
determine a common set of policies among the partners, and insure that
all the communication requirements for the coalition can be met by all
the partners.  If the requirements cannot be met, they must be flagged
so an administrator can fix them before they interfere with the

Condell                                                         [page 3]

Internet Draft                MSME and IPSP                November 2001

Additionally, once the policies are in place, partners will want to
ensure that the policies are being used and enforced correctly.  Each
partner requires some level of monitoring to verify that it and other
partners are correctly enforcing the agreed upon policies.  How much
information may be obtained from monitoring will depend upon where
monitoring probes are allowed to be placed, which, in turn, affects
what information is available to monitor.

2.2 Overview

The MSME architecture consists of a high-level policy description
language; protocols for transferring the policy between partners; and
algorithms for ensuring the internal consistency of the high-level
policy description, resolving the policies between partners, and
verifying that the resolved policy does not violate the policies of a
particular partner.

When a coalition forms, membership changes, or one or more partners'
policies change, each partner creates a policy level agreement (PLA).
The PLA contains high-level policy requirements and mappings from the
abstract requirements to concrete implementations that the partner
supports.  Each partner checks that their PLA is self consistent and
that their local policies do not conflict with the high-level
requirements.  If there is a conflict or inconsistency, human
intervention is required to correct the problem.

The partners then exchange their PLAs.  Depending on the coalition,
partners may either exchange their PLAs with each other, or they may
hand them to a central server.  The exchange must be authenticated and
may require confidentiality.  Each partner, or a central server, as
agreed to by the coalition, must then resolve the PLAs.  The
resolution process determines which concrete policies the coalition
members agree to use to implement the policies specified in their
PLAs.  PLAs may be resolved by subsets of the coalition partners to
allow services among the subsets that do not involve the rest of the
coalition.  The resolution process produces a resolved PLA (RPLA).
The RPLA will have to be distributed to each partner if the resolution
process was centralized.

When each partner receives or computes the RPLA, it must reconcile it
with its local policies to verify that it does not violate its local
policy.  Additionally, the reconcilliation process can detect policy
rules that were not included in the RPLA because other coalition
members could not support them and may require manual intervention if
those communications must be supported.  If the resolution process is
executed properly and no partner misbehaves, the reconciliation
process should not detect any rule violations, however, it is a
necessary step to ensure there were no problems.

The dynamic coalition environment will produce a very complex set of
policies, so it will be difficult to determine a priori if the correct
policies were negotiated.  Monitoring can help determine if the
correct policies are being used and that policy enforcement points are

Condell                                                         [page 4]

Internet Draft                MSME and IPSP                November 2001

using them correctly.  The power of monitoring will be limited by
where partners will allow monitoring probes to be placed, both
physically and because most communications will be encrypted.

Partners may have to take into account that policy may be defined by
different people at different levels.  The high-level policy may be
defined by someone who established the coalition and its communication
requirements, but does not know how the requirements get implemented.
Similarly, bindings may be created by someone who understands the
low-level policies, but did not determine the high-level policy.

2.3 Requirements

This section will discuss some of the main requirements for the MSME

 * MSME MUST allow each partner to internally implement its own policy
   languages, policy storage and policy distribution mechanisms.
   However, MSME MAY impose requirements on them.

 * MSME MUST provide a mechanism for partners to exchange their
   policies and to resolve policy information using a common protocol.
   The policies must be securely communicated, including
   authentication and integrity checks.  MSME SHOULD have a mechanism
   by which coalition partners can associate security policy rules
   with specific coalition partners.

 * MSME SHOULD support both private peering (sub-coalitions) and
   partial (or abstracted) sharing of internal policy information.
   Some partners may want to keep portions of their policy private
   from other partners.  Obviously, those rules of their policy
   relevant to the policy resolution need to be released to other

 * MSME MUST provide a way to determine whether a specific
   communication is permitted by the current policy agreement between
   coalition partners before the communication is attempted.

 * MSME MUST provide a way to establish and identify, at any
   time, the set of authorized coalition partners and the entities
   that they have authorized to engage in coalition activities.

 * MSME MUST support security services implemented by multiple
   security protocols (e.g., IPsec, TLS), and compositions thereof.
   Therefore, MSME SHOULD support a security abstraction layer that
   can map (high-level) policy intent to different (low-level)
   security policy data models. Both the representation (language) and
   exchange (protocol) of this abstraction must be supported.

 * MSME MUST provide a way to monitor policies while they are in use
   to confirm that the correct policies are installed in the
   enforcement points, that they are being applied correctly, and that

Condell                                                         [page 5]

Internet Draft                MSME and IPSP                November 2001

   the policy decision points (PDPs) and policy enforcement points
   (PEAs) are behaving correctly.  There must be a way to monitor both
   the coalition policies and the local policies of the PDPs and PEAs.

2.4 Architecture

Figure 1 illustrates the components of the MSME system.  The
partner-dependent components are not defined as part of the MSME
system, however, the system interacts with those components and may
impose requirements on them.  Intra-partner components represent
components fully contained within a partner.  Inter-partner components
which are not partner-dependent must exist as part of the MSME system,
but don't have to directly interoperate with other partners.  Other
components are part of the MSME system and must be interoperable
between partners.

 |  Partner-dependent  |                     MSME                     |
 |.........................................                           |
 |.               Intra-partner           .                           |
 |.  +--------------+  | +--------------+ .           +-------------+ |
 |.  |    PMT/      |  | |     Policy   | .           |  Resolution | |
 |.  |Administrator |  | |  Compilation | .           |             | |
 |.  +--------------+  | +--------------+ .           +-------------+ |
 |.                    |                  .                           |
 |.  +--------------+  |           +-------------+    +-------------+ |
 |.  |    Policy    |  |           |Policy Level |    |   Exchange  | |
 |.  |   Language   |  |           |  Agreement  |    |   Protocol  | |
 |.  +--------------+  |           +-------------+    +-------------+ |
 |.                    |                  .                           |
 |.  +--------------+  | +--------------+ .                           |
 |.  | Local Policy |  | |Reconciliation| .                           |
 |.  | Repositories |  | |              | .                           |
 |.  +--------------+  | +--------------+ .                           |
 |.                    |                  .                           |
 |.  +--------------+  |           +-------------+                    |
 |.  |     PEPs     |  |           |  Monitoring |                    |
 |.  |              |  |           |             |                    |
 |.  +--------------+  |           +-------------+                    |
 |.........................................                           |

              Figure 1.   MSME Architecture Components

The remainder of this section will describe the components and their
interactions in some detail and will provide a description of data
flow through the system to illustrate how the system is designed to

Condell                                                         [page 6]

Internet Draft                MSME and IPSP                November 2001

2.4.1 MSME Components Policy Level Agreement

Policy level agreements (PLAs) are the means by which each of the
partners in a coalition expresses and exchanges its policy.  The PLAs
are expressed in a common language (PLAL) to facilitate exchange and

PLAs include the following information:
 * The identity of the coalition to which a PLA applies
 * The members of the coalition
 * The member that created the PLA
 * Versioning information for the PLA
 * Abstract policy rules using arbitrary identifiers chosen by each
   partner Some identifiers may be agreed upon to be globally unique.
 * Bindings that map the identifiers to concrete assets or security

Policy rules are defined at an abstract level, with conditions
expressed in terms of abstract entities or entity groups and actions
expressed in terms of security services and abstract mechanisms.  The
services and mechanisms are a superset of those described in

Each abstract entity, service, or mechanism includes a textual name
which identifies one or more bindings which map the abstract part of
the rule to a concrete implementation (e.g., an IP address, or a
security protocol and its acceptable algorithms).  Multiple bindings
may have the same name, since the required service may be implemented
in multiple security contexts (e.g., IPsec, TLS).

Some binding names may be globally recognized by the coalition
partners so that a partner may refer to other partners' entities
without having concrete bindings.  For example, partner 1 can have an
abstract policy that requires ESP to be used between its gateways and
partner 2's gateways, without having to know the IP addresses of
partner 2's gateways when defining its policy.  Those addresses will
be supplied by partner 2 in its PLA and will be bound to partner 1's
policy during the resolution process. Policy Compilation

Compilation is the process by which each partner generates a PLA from
its abstract policy rules.  The compilation process takes the set of
abstract policy rules and a database of bindings, assembles the
bindings needed to support the policy rules, confirms all the bindings
are available, other than those supposed to be supplied by other
partners, and checks to make sure that the policy rules are
self-consistent (i.e., the policies don't contain any contradictions).
The output of the process is either a complete PLA that may be

Condell                                                         [page 7]

Internet Draft                MSME and IPSP                November 2001

exchanged with other partners, or errors indicating that the PLA is
not consistent or complete and must be fixed by the policy
administrator. Resolution

Resolution is the process of combining all (or some subset) of the
partners' PLAs into a single PLA containing rules consistent with all
of their policies. The 'combined' PLA is known as a resolved PLA

Resolution combines two or more member PLAs in an attempt to generate
a single RPLA that is consistent with all of the policies expressed in
the individual PLAs. It identifies the security mechanisms supported
by all the parties required for each mission-related policy rule.  It
also identifies the cases where all partners do not support a common
security mechanism for a policy rule, an indication that a
communication that cannot be successfully initiated due to current
policies, thus allowing the policy administrators to address any
potential deficiencies in their respective policy specifications. This
`diagnostic' use of resolution is an important function in real world

Resolution may be performed in a centralized manner, using a single
resolver for the coalition; in a decentralized manner, with each
member performing an independent resolution of the same set of PLAs;
or in some intermediate way. The MSME resolution mechanism should be
designed to support decentralized operations, as centralized
operations can be viewed as a special case of the more general
decentralized resolution architecture. Exchange Protocol

The PLA exchange protocol is a common protocol is used for the
exchange of PLAs and RPLAs, and for monitoring the status of the
resolution process.

The exchange of PLAs and RPLAs between all partners requires some
exchange mechanism. Although nothing fundamentally precludes partners
from implementing differing exchange mechanisms on a bilateral or
multilateral basis, it is useful to consider a common exchange
protocol. Even if such a protocol is not universally adopted, its
architectural framework is helpful in ensuring that whatever exchange
mechanisms are implemented sufficiently support the requirements of
the MSME system.

The protocol's primary aim is to support the resolution process, by
allowing PLAs and RPLAs to be exchanged, and by allowing resolution
status information to be conveyed. The protocol should support the
following operations:

Condell                                                         [page 8]

Internet Draft                MSME and IPSP                November 2001

 * Exchange of PLAs between partners, or from partners to a central
   resolution authority.

 * Exchange of RPLAs: In a centralized resolution architecture,
   centrally resolved RPLAs will have to be propagated to coalition
   partners. In a totally decentralized environment, partners may
   still need to request each others' RPLAs for reconciliation

 * Co-ordination/synchronization of the resolution process: The
   resolution process is essentially asynchronous, regardless of
   whether it is performed centrally or in a distributed manner. Some
   mechanism is required to inform entities within the coalition of
   the current status of the resolution process. This may, for
   example, include a means for requesting a new round of resolution,
   or for of informing partners of the (un-)successful completion of a
   resolution process.

The protocol must authenticate the source of the PLA and optionally
provide data confidentiality for the exchange. Reconciliation

Once an RPLA has been generated, a reconciliation process must occur
so that each partner may verify that the RPLA is correct and that it
does not conflict with its internal policy rules.

A coalition partner cannot assume that any resolution that it did not
perform itself will yield a correct RPLA.  The partner must confirm
that: no policy rules introduced to the coalition by the partner in
its PLA have been removed, no bindings have been defined in a manner
that conflicts with any of the definitions introduced by the partner
in its PLA, and no policy rules have been introduced in the RPLA that
conflict with any of the partner's policy rules in its PLA.
Additionally, a partner may use reconcilliation to identify places
where resolution occurred correctly, but did not achieve the desired
result, and may have to be adjusted by administrative intervention.
For example, it could detect policy rules that are not implementable
because no common security mechanisms exists with other partners, or
cases where a weaker, though still acceptable, mechanism was selected
instead of a stronger, more desireable one.

An incorrect RPLA may be generated for a number of reasons, including
incorrect implementation of the resolution algorithm, PLA
synchronization issues, or malicious intent by one of the coalition
partners. In order to ensure the correctness of a new RPLA that it has
received, a partner should evaluate its freshness and correctness before
provisioning it.

Condell                                                         [page 9]

Internet Draft                MSME and IPSP                November 2001 Monitoring

Monitoring ensures that RPLAs are being correctly generated and
enforced.  A number of elements of the MSME system will be involved in
the process of monitoring that RPLAs are correctly implemented. The
cornerstone of the monitoring aspect of MSME is the reconciliation
process, which provides each partner with an essential sanity check
that the MSME system is generating correct policy. The versioning
mechanism provided by the PLAL provides a means to allow partners to
compare the versions of the RPLAs that they are using to ensure that
they are operating with consistent policies. Version checking is
further facilitated by the MSME PLA exchange protocol which allows
partners to query the status of the resolution process, and to
obtain the currently active RPLA.

Additional monitoring tasks may include verifying that enforcement
points are being configured with the correct policies and that they
are using them correctly.  This monitoring is necessarily limited
since most communications are likely to be encrypted and a partner
will generally only have access to the unencrypted messages on its own

2.4.2 Partner-Dependent Components Policy Management Tool

The policy management tool (PMT) provides a user interface to the MSME
system.  It provides an administrator the means to enter abstract
policies and bindings into the system.  The PMT also must be able to
read the policies in a PLA or local repository and display the results
to the administrator.  The administrator should be able to modify
these policies.

Policy management goes beyond just creating and viewing policies.  The
PMT should provide an interface to other functions.  For example, it
should be able to initiate the policy compilation process and display
the results, including any warnings or errors that it produces, so the
administrator can correct them. It may also be responsible for
initiating resolution, or at least for sending PLAs to a central
resolution point. Policy Language

The choice of languages that a partner uses for defining its internal
policies are up to the partner, however there must be tools to
translate those languages to and from the PLA language in order to
create bindings and to translate RPLAs into the language(s) that the
local managment system uses to provision policy enforcement points.

Condell                                                        [page 10]

Internet Draft                MSME and IPSP                November 2001 Local Policy Repositories

Each partner maintains collections of information relating to their
existing internal policy environments. The nature of these
repositories is highly partner-specific.  There may be databases
specifically for MSME data, or they may support both MSME and internal
policy management systems and provide the link between the two.  This
section discusses what information the repositories must be able to
provide MSME.

The following information should be available:

 * Asset data: Mappings between high-level assets and their
   composition in terms of low-level assets.

 * Security service to mechanism mappings: Whenever a high-level
   policy requires a security service, this repository is consulted to
   determine what mechanisms are available to implement that service.

 * Local mechanism-specific security policy repositories

 * Optionally, repositories of global data

The information in the local repositories is used in the generation of
bindings. For each asset specified in the high-level policy, the
corresponding low-level endpoints are determined from the asset
repository.  Then for each low-level mechanism, the relevant
mechanism-specific repository is consulted to extract the information
required for policy resolution.

The local policy repository for each context is consulted to determine
which bindings are valid (from local context-specific policy
considerations) and may be included in the PLA.  This process must
also be repeated with the resolved PLAs to ensure that the RPLA does
not violate any local policy constraints. Policy Enforcement Points

Once a partner has received and successfully reconciled an RPLA, it
is up to the policy enforcement points to enforce the resolved
policies.  This may require the partner to provision new
policy rules to its internal PEPs.

This will typically involve reprovisioning policy management and
enforcement agents within the partner's security domain. While this
provisioning is outside the scope of the MSME system, any MSME
implementation will have to consider the interaction between the RPLA
validation and acceptance mechanism, which is part of the core MSME
system, and the policy provisioning system, which is outside of it.

The nature of this link between MSME and policy provisioning
mechanisms is highly dependent on the nature of the specific
provisioning mechanisms in use. Conceptually, it would typically

Condell                                                        [page 11]

Internet Draft                MSME and IPSP                November 2001

involve the inverse of the process of extracting information
out of the local policy repositories. The new policy can then be
distributed by mechanisms such as those being developed by the
IPSP working group.

2.4.3 Data Flow

This section describes how the components of the MSME system work
together to provide policy management for dynamic coalitions.

             +------------+          +------------+
             |  Abstract  |          |  Bindings  |
             |  Policies  |          |            |
             +------------+          +------------+
                        |              |
                        v              v
                     | Policy Compilation |
                +-----+     +-----+      +-----+
                | PLA |     | PLA |      | PLA |
                +-----+     +-----+      +-----+
                     \         |          /
                      v        v         v
                     | Policy Resolution  |
                            |RPLA |
                     |   Reconciliation   |
                     | Provision policies |

                    Figure 2: MSME Data Flow

Policies are created by policy administrators.  The people who create
the abstract policies and the bindings do not have to be the same,
since the people who understand the coalition's communication
requirements are not necessarily the same as those who understand the
concrete policies that the partner supports.

Condell                                                        [page 12]

Internet Draft                MSME and IPSP                November 2001

The MSME resolution process may be initiated for several reasons,
including a change in policy, abstract or concrete, or a change in the
coalition membership.

A partner must first generate a new PLA when a resolution exchange is
initiated and a partner determines that its policies (including
abstract rules, bindings, and contexts) have changed.  If its policies
have not changed then its previous PLA may be reused.  To generate a
new PLA, the compilation process is invoked.  It takes the abstract
policy rules and the database of bindings, extracts the necessary
bindings and produces a consistent PLA, if possible.  If there were
errors, they must be corrected manually by the policy administrators.

The PLA is sent, using the exchange protocol to one or more policy
resolvers that will be creating RPLAs. The policy resolver creates
an RPLA that satisfies the requirements of all the RPLAs, if possible.
If it is not possible, then an error is returned and the partners need
to manually correct their PLAs, if they wish, and attempt the resolution
again.  Once the RPLA is produced it is sent to the partners, as needed.

The first step a partner performs upon receiving a new RPLA is
reconciliation to ensure that the RPLA is in fact consistent with its
policies.  If reconciliation fails, the partner should correct the
problem, if possible, or report it to the other partners which may
have to make corrections to their policies to correct the problems.

If reconciliation succeeds, the partner should commence enforcement of
the RPLA. Again, the details of the operations required to enforce the
RPLA are partner-specific, but the overall principles are the same:
new concrete rules may need to be provisioned at enforcement points to
accommodate coalition policy. This necessarily implies a narrowing of
existing policy, since a broadening would require human negotiation
(if resolution failed, for example, one or more partners might agree
to relax their local policy restrictions to allow resolution to

3. Ideas for IPSP

The MSME system is attempting to solve a different problem than is
currently being addressed by the IPSP working group so while its
architecture contains many similar elements, they are instantiated
differently. However, since it is also working in the security policy
management domain it may provide some insights that can improve the
final output of the IPSP WG, especially in the area of defining and
generalizing policy rules.  This section explores some of the areas
where MSME's work may provide some insights into improving the IPSP

3.1 Late Name Binding

Early IPSP working group drafts, now expired, such as the "Security
Policy Specification Language" (draft-ietf-ipsp-spsl-00.txt) and
"Security Policy Protocol" (draft-ietf-ipsp-spp-00.txt) included the

Condell                                                        [page 13]

Internet Draft                MSME and IPSP                November 2001

beginnings of an idea that some hosts should not be specified by IP
address, but could be addressed in a more general way.  This allowed
policies, which were configured in the security servers, to be defined
in such a way that it did not need to specify an IP address for hosts
or gateways it may not know the IP address, or even if such a gateway
might exist.  For example, those two drafts allowed policies to use
some defined strings such as "host", "remote-sg", or "local-sg" to
refer to such entities.

While it was an interesting idea, it was poorly developed and, as
described in those drafts, not completely implementable.  While the
"host" tag could easily be interpreted as either the source or
destination of the communication, none of the others could be reliably
translated into a specific host or gateway.  MSME provides some
insights into how to design this feature in an implementable fashion
and possibly ways to expand on the idea.  In particular adding the
idea of binding a name to one or more assets could add this

The protocol and language could continue to use standardized strings
to refer to particular types of hosts as those drafts indicated.
However, the security servers would use bindings as part of their
local policies to map the strings to the appropriate network entities.
Their mapping could even be different for different policy rules so that
a communication being negotiated for host A could have a different
mapping than a communication being negotiate for host B, for example.

To illustrate this, let's look at the following example:

   Host A ---- GW A ---- network ---- GW B -------------- Host B
                                               GW C ----- Host C

Host A initiates a request to discover the policy requirements for a
communication between itself and Host B.  GW A has a policy that
requires a tunnel for that communication between itself and any
gateway that is authoritative over Hosts B or C (indicated by the
string "remote-sg").  GW B (or its policy server) has a binding that
maps the string "remote-sg" to GW B for this communication so GW A
then learns as part of the resolution that it needs to have a tunnel
with GW B.

If Host A then wants to communicate with Host C and the policies are
the same, GW B can have a binding that maps "remote-sg" to GWs B and C
so GW A can tunnel to either of them.

We can extend this concept beyond a few standardized strings that all
policy domains can implement.  If we allow arbitrary names to be used
(or at least a private namespace), then security policy domains can
privately agree upon names to use to abstract portions of their policy

For example, if companies X and Y need to facilitate communications
between their engineering departments, they could agree upon "x_eng"

Condell                                                        [page 14]

Internet Draft                MSME and IPSP                November 2001

to be a name for X's engineering department and "y_eng" for Y's.  X
can have a binding that maps "x_eng" to the necessary hosts in its
engineering department and Y can do similarly.  Now X can create
policies in terms of "y_eng" instead of a list of all the IP addresses
in Y's engineering department.

This has a couple of advantages.  Now the hosts in Y's engineering
department can change without Y having to inform X, since all Y needs
to do is change its binding and X's policy is still valid since the
name does not change.  Additionally, it can help reduce the amount of
network information that each company needs to expose to the other.

While it may be possible to achieve some of this abstraction by using
wildcarded DNS names (e.g., * in policy rules, the binding
mechanism is much more powerful and general.  Not only may it map from
a name to one or more hosts, but it can map from a name to one or more
conditions that are required (e.g., port, protocol, etc).  For
example, company Y could have different bindings for its HTTP and FTP
servers, even if they are the same hosts, with the same host names,
since each binding would map the binding name to a set of hosts and a
port number.

3.2 Generalization

Currently, IPSP is looking exclusively at providing security policy
management for the IPsec security domain.  If this can be accomplished
in a more general manner that allows the solution to be applied to
other security domains, it would increase the usefulness of the work,
without greatly increasing its scope.

One possible means of generalizing the IPSP work is to allow policy
rules to express multiple security domain options in one policy rule
by using bindings.

In order to accomplish this, policy rules must be able to be expressed
using names for the source and destination of the condition and the
action.  For example, Company Y could have the policy:

   src y_eng dst x_eng -> strong_authentication

Now, suppose both companies X and Y are willing to use either IPsec or
TLS to protect the communication.  "y_eng," "x_eng," and
"strong_authentication" can refer to bindings that have both IPsec and
TLS mappings which can be sent along with the policy rule as part of
the policy discovery phase of IPSP's protocol.  "y_eng" may map to a
set of IP addresses and port numbers in the IPsec context.  In the TLS
context it may map to a set of user names.  "strong_authentication"
may also map to the required mechanisms in each context such as an ESP
proposal or a TLS mac algorithm.

Company X can then have a similar policy:

   src y_eng dst x_eng -> xs_authentication

Condell                                                        [page 15]

Internet Draft                MSME and IPSP                November 2001

X will have mappings for "x_eng" and "xs_authentication" in a similar
manor to Y.  Note that "x_eng" and "y_eng" can be pre-agreed upon names
between the two companies as described above.  "strong_authentication"
and "xs_authentication" are not, however.

Resolution of the two policies is performed at the mechanism level, as
would be done if the bindings did not exist.  Only bindings in the
same context can be resolved and each context is resolved
independently.  This allows the resolution to succeed if at least one
condition/action pair in one context can successfully resolve.  If
multiple contexts successfully resolve, then both may be returned as
the answer.

4. Acknowledgments

The author thanks the MSME team, Luis Sanchez, Charlie Lynn, Geva
Patz, Alex Colvin, David Waitzman, and Rajesh Krishnan for their work
developing the MSME system and reviewing this draft.

5. References

[Bra97] Bradner, S., "Key Words for use in RFCs to indicate
    Requirement Levels", RFC2119, March 1997.

[TERM] A. Westerinen, J. Schnizlein, J. Strassner, M. Scherling,
    B. Quinn, S. Herzog, A. Huynh, M. Carlson, J. Perry, and
    S. Waldbusser, Terminology for Policy-Based Management
    draft-ietf-policy-terminology-04.txt (approved for RFC)

[iso7498-2] ISO. "Information Processing Systems - Open Systems
    Interconnection - Basic Reference Model - part 2: Security
    Architecture," first edition, International Standard ISO 7298-2,
    ISP, February 1989.

Condell                                                        [page 16]

Internet Draft                MSME and IPSP                November 2001


   The views and specification here are those of the authors and are
   not necessarily those of their employers.  The authors and their
   employers specifically disclaim responsibility for any problems
   arising from correct or incorrect implementation or use of this

   Copyright (C) The Internet Society (2001).  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

Author Information

   Matthew N. Condell
   BBN Technologies
   10 Moulton Street
   Cambridge, MA  02138
   Telephone: +1 (617) 873-6203

Condell                                                        [page 17]