wdiff draft-schoenw-snmp-tlsm-01.txt draft-schoenw-snmp-tlsm-02.txt

Network Working Group D. Harrington
Internet-Draft Independent
Expires: April 1, November 27, 2005 J. Schoenwaelder
International University Bremen
October 2004
May 26, 2005

Transport Mapping Security Model (TMSM) for the Simple Network
Management Protocol version 3 (SNMPv3)
draft-schoenw-snmp-tlsm-01.txt
draft-schoenw-snmp-tlsm-02.txt

Status of this Memo

This document is an Internet-Draft and is subject to all provisions
of section 3 of RFC 3667.

By submitting this Internet-Draft, each author represents that any
applicable patent or other IPR claims of which he or she is aware
have been or will be disclosed, and any of which he or she become becomes
aware will be disclosed, in accordance with
RFC 3668. Section 6 of BCP 79.

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. Internet-
Drafts.

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
http://www.ietf.org/ietf/1id-abstracts.txt.

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.

This Internet-Draft will expire on April 1, November 27, 2005.

Abstract

This document describes a Transport Mapping Security Model (TMSM) for
the Simple Network Management Protocol (SNMP) architecture defined in
RFC3411. At this stage, this document describes a framework, not a
protocol. It does not provide a complete solution - it rather
identifies and discusses some key aspects that need discussion and
future work.

Table of Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Requirements of a Transport Mapping Security Model . . . . . . 4 5
3.1 Security Requirements . . . . . . . . . . . . . . . . . . 4
3.2 Architectural Modularity 5
3.1.1 Security Protocol Requirements . . . . . . . . . . . . 5
3.3 Passing messages between Dispatchers
3.1.2 Session Requirements . . . . . . . . . . . . . 5
3.4 Security Parameter Passing Requirement . . . . 6
3.2 Architectural Modularity Requirements . . . . . . . . . . 6
3.4.1 Using an ASI
3.2.1 USM and the RFC3411 Architecture . . . . . . . . . . . 9
3.2.2 TMSM and the RFC3411 Architecture . . . . . . . . . . 6
3.4.2 Using a cache 10
3.3 Passing Messages between Subsystems . . . . . . . . . . . 11
3.4 Security Parameter Passing Requirement . . . . . . . . . . 12
3.4.1 Define an Abstract Service Iinterface . 6
3.4.3 . . . . . . . 13
3.4.2 Using an encapsulating header . . . . . . . . . . . . 7 13
3.4.3 Modifying Existing Fields in an SNMP Message . . . . . 13
3.4.4 Using existing fields in a message cache . . . . . . . . . . 7
3.5 Access Control Requirements . . . . . . . . . . 14
3.5 Architectural Requirements for Access Control . . . . . 8 . 14
3.5.1 Architectural securityName Binding Requirement . . . . 8
4. Fields in the SNMPv3 message . . . . . . . . . . . . . 14
3.5.2 Separation of Authentication and Authorization . . . . 15
4. Integration with the SNMPv3 message format . . . . . . . . . 8 . 16
4.1 msgVersion . . . . . . . . . . . . . . . . . . . . . . . . 8 16
4.2 msgGlobalData . . . . . . . . . . . . . . . . . . . . . . 9 16
4.3 securityLevel and msgFlags . . . . . . . . . . . . . . . . 9 17
4.4 The tmStateReference for Passing Security Parameters . . . 11 18
4.5 securityStateReference Cached Security Data . . . . . . . 11 18
4.5.1 Prepare an Outgoing SNMP Message . . . . . . . . . . . 12 19
4.5.2 Prepare Data Elements from an Incoming SNMP Message . 12 20
4.6 Notifications . . . . . . . . . . . . . . . . . . . . . . 13 20
5. Transport Mapping Security Model Samples . . . . . . . . . . . 13 21
5.1 TLS/TCP Transport Mapping Security Model . . . . . . . . . 13 21
5.1.1 tmStateReference for TLS . . . . . . . . . . . . . . . 13 21
5.1.2 MP portion MPSP for TLS TM-Security Model . . . . . . . . . 14 . . . 22
5.1.3 MIB Module for TLS Security . . . . . . . . . . . . . 14 22
5.2 DTLS/UDP Transport Mapping Security Model . . . . . . . . 14 22
5.2.1 tmStateReference for DTLS . . . . . . . . . . . . . . 15 23
5.3 SASL Transport Mapping Security Model . . . . . . . . . . 16 24
5.3.1 tmStateReference for SASL DIGEST-MD5 . . . . . . . . 16 24
6. Security Considerations . . . . . . . . . . . . . . . . . . . 25
7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 17
7. 25
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 17
7.1 25
8.1 Normative References . . . . . . . . . . . . . . . . . . . . 17
7.2 25
8.2 Informative References . . . . . . . . . . . . . . . . . . . 18 26
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 18 27
A. Message security versus session security Questions about msgFlags: . . . . . . . . . . . 19 . . . . . . . 27
A.1 msgFlags versus actual security . . . . . . . . . . . . . 19 27
A.2 Message security versus session security . . . . . . . . . 19 29
Intellectual Property and Copyright Statements . . . . . . . . 21 30

1. Introduction

There are multiple ways to secure one's home or business, but they
largely boil down to a continuum of alternatives. Let's consider
three general approaches. In the first approach, an individual/
company individual could
buy a gun, learn to use it, and sit on your front porch waiting for
intruders. In the second approach, one could hire an employee with a
gun, schedule the employee, position the employee to guard what you
want protected, hire a second guard to cover if the first gets sick,
and so on. In the third approach, you could hire a security company,
tell them what you want protected, and they could hire employees,
train them, buy the guns, position the guards, schedule the guards,
send a replacement when a guard cannot make it, etc., thus providing
the security you want, with no significant effort on your part other
than identifying requirements and verifying the quality of the
service being provided.

The User-based Security Model (USM) as defined in [RFC3414] largely
uses the first approach - it provides its own security. It utilizes
existing mechanisms (MD5=the gun), but provides all the coordination.
USM provides for the authentication of a principal, message
encryption, data integrity checking, timeliness checking, etc.

USM was designed to be independent of other existing security
infrastructures. USM therefore requires a separate user and key
management infrastructure. Operators have reported that deploying
another user and key management infrastructure in order to use SNMPv3
is a reason for not deploying SNMPv3 at this point in time. It is
possible but difficult to define external mechanisms that handle the
distribution of keys for use by the USM approach.

A solution based on the second approach might use a USM-compliant
architecture, but replace the authentication mechanism with an
external mechanism, such as RADIUS, to provide the authentication
service. It might be possible to utilize an external protocol to
encrypt a message, to check timeliness, to check data integrity, etc.
It is difficult to cobble together a number of subcontracted services
and coordinate them however, because it is difficult to build solid
security bindings between the various services, and potential for
gaps in the security is significant.

A solution based on the third approach might utilize one or more
lower-layer security mechanisms to provide the message-oriented
security services required. These would include authentication of
the sender, encryption, timeliness checking, and data integrity
checking. There are a number of IETF standards available or in
development to address these problems at lower layers, frequently at
the transport layer. A solution based on this approach might also
utilize a "transport" "transport application" that is actually another
application operating at the application layer, such as SSH [SSHauth]

This document proposes a Transport Mapping Security Model (TMSM), as
an extension of the SNMPv3 architecture, that would allow security to
be provided by an external protocol connected to the SNMP engine
through an SNMP transport-mapping. Such a TMSM would then enable the
use of existing security mechanisms such as (TLS) [RFC2246], Kerberos
[RFC1510] or SASL [RFC2222] within the SNMPv3 architecture.

As pointed out in the EUSM proposal [EUSM], it is desirable to use
mechanisms that could "unify the approach for administrative security
for SNMPv3 and CLI" and other management interfaces. The use of
security services provided by lower layers or other applications is
the approach commonly used for the CLI, and is the approach being
proposed for NETCONF

This document provides describes the motivation for leveraging transport layer
security mechanisms for secure SNMP communication, identifies some
key issues and provides some proposals for design choices that may be
made to provide a workable solution that meets operational
requirements and fits into the SNMP architecture defined in [RFC3411]

2. Motivation

There are a number of Internet security protocols and mechanisms that
are in wide spread use. Many of them try to provide a generic
infrastructure to be used by many different application layer
protocols. The motivation behind TMSM is to leverage these
protocols where it seems useful.

There are a number of challenges to be addressed to map the security
provided by a secure transport into the SNMP architecture so that
SNMP continues to work without any surprises. These are discussed in
detail below.

Some points requiring further WG research and discussion are
identified by [todo] markers in the text.

3. Requirements of a Transport Mapping Security Model

3.1 Security Requirements

Transport mapping security protocols SHOULD ideally provide the
protection against the following message-oriented threats [RFC3411]:

1. modification of information
2. masquerade
3. message stream modification
4. disclosure

According to [RFC3411], it i
s is not required to protect against denial
of service or traffic analysis.

3.1.1 Security Protocol Requirements

There are a number of standard protocols that could be proposed as
possible solutions within the TMSM framework. Some factors should be
considered when selecting a protocol for use within this framework.

Using a protocol in a manner for which it was not designed has
numerous problems. The advertised security characteristics of a
protocol may depend on its being used as designed; when used in other
ways, it may not deliver the expected security characteristics. It
is recommended that any proposed model include a discussion of the
applicability statement of the protocols to be used.

A protocol used for the TMSM framework should ideally require no
modifications to the protocol. Modifying the protocol may change its
security characteristics in ways that would impact other existing
usages. If a change is necessary, the change should be an extension
that has no impact on the existing usages. It is recommended that
any proposed model include a discussion of potential impact on other
usages of the protocol.

It has been a long-standing requirement that SNMP be able to work
when the network is unstable, to enable network troubleshooting and
repair. The UDP approach has been considered to meet that need well,
with an assumption that getting small messages through, even if out
of order, is better than gettting no messages through. There has
been a long debate about whether UDP actually offers better support
than TCP when the underlying IP or lower layers are unstable. There
has been recent discussion of whether operators actually use SNMP to
troubleshoot and repair unstable networks.

There has been discussion of ways SNMP could be extended to better
support management/monitoring needs when a network is running just
fine. Use of a TCP transport, for example, could enable larger
message sizes and more efficient table retrievals.

TMSM models MUST be able to coexist with other protocol models, and
may be designed to utilize either TCP or UDP, depending on the
transport.

3.1.2 Session Requirements

Sessions are not part of RFC3411 architecture, but are considered
desirable because the cost of authentication can be amortized over
potentially many transactions.

For transports that utilize sessions, a session should have a single
user and security level associated with it. If an exchange between
communicating engines would require a different security level or
would be on behalf of a different user, then another session would be
needed. An immediate consequence of this is that implementations
should be able to maintain some reasonable number of concurrent
sessions.

3.2 Architectural Modularity Requirements

[RFC3411] section 3 describes a modular architecture to allow the
evolution of the SNMP protocol standards over time. time, and to minimize
side effects between subsystems when changes are made. This
architecture includes a Security Subsystem which is responsible for
realizing security services.

Transport mapping security is

In SNMPv2, there were many problems of side effects between
subsystems caused by its very nature a security layer
which is plugged the manipulation of MIB objects, especially
those related to authentication and authorization, because many of
the parameters were stored in between shared MIB objects, and different
models and protocols could assign different values to the transport layer objects.
Contributors assumed slightly different shades of meaning depending
on the models and protocols being used. As the dispatcher.
Conceptually, transport mapping security shared MIB module
design was modified to accommodate a specific model, other models will be called
which used the same MIB objects were broken.

ASIs were developed to pass model-independent parameters. The models
were required to translate from
within their model-dependent formats into a
model-independent format, defined using model-independent semantics,
which would not impact other models.

Parameters have been provided in the Transport Mapping portion ASIs to pass model-independent
information about the authentication that has been provided. These
parameters include a model-independent identifier of an SNMP engine, or will be
positioned between the transport mapping subsystem security
"principal", the security model used to perform the authentication,
and which SNMP-specific security features were applied to the
dispatcher. message
(authentication and/or privacy).

The design of a transport mapping security model must abide the goals
of the RFC3411 architecture, section 1. architecture. To that end, this transport mapping
security model proposal focuses on a modular subsystem that can be
advanced through the standards process independently of other
proposals, and independent of other subsystems as much as possible.

This subsystem is designed as an architectural extension that permits

There has been some discussion of maintaining multiple tunnels or
sessions for different transport mapping security protocols levels or for different
applications.The ability to be "plugged into"
this subsystem, have an application select different
sessions or connections on a per-message basis would likely require a
modification to support supplemental transport mapping security
models in addition the SNMP architecture to those described here. provide new ASIs, which is
out of scope for this document.

IETF standards typically require one mandatory-to-implement solution,
with the capability of adding new security mechanisms in the future.
Any transport mapping security model should define one
minimum-compliance minimum-
compliance mechanism, preferably one which is already widely deployed
within the transport layer security protocol used.

This

The TMSM subsystem is designed as an architectural extension that
permits additional transport security protocols to be "plugged into"
the RFC3411 architecture, supported by corresponding transport-
security-aware transport mapping models.

The RFC3411 architecture, and the USM approach, assume that a
security model is illustrated called by a message-processing model and will
perform multiple security functions. The TMSM approach performs
similar functions but performs them in different places within the following diagram,
archtitecture, so we need to distinguish the two locations for
security processing.

Transport mapping security is by its very nature a security layer
which is plugged into the RFC3411 architecture between the transport
layer and the message dispatcher. Conceptually, transport mapping
security processing will be called from within the Transport Mapping
functionality of an SNMP engine dispatcher to perform the translation
of transport security parameters to/from security-model-independent
parameters. This transport mapping security processor will be
referred to in this document as TMSP.

Additional functionality may be performed as part of the message
processing function, i.e. in the security subsystem of the RFC3411
architecture. This document will refer to message processor's
security processor as the MPSP.

Thus a modified version TMSM is composed of both a TPSP and an MPSP.

+------------------------------+
| Network |
+------------------------------+
^ ^ ^
| | |
v v v
+-----+ +-----+ +-------+
| UDP | | TCP | . . . | other |
+-----+ +-----+ +-------+
^ ^ ^
| | |
v v v
+------+ +-----+ +-------+
| DTLS | | TLS | . . . | other |
+------+ +-----+ +-------+ (traditional SNMP agent)
+-------------------------------------------------------------------+
| ^ |
| | |
| Dispatcher v |
| +-------------------+ |
| | Transport | +--------------------+ |
| | Mapping |<---> | Transport Mapping | |
| | (e.g., RFC 3417) | | Security Processor | |
| | | +--------------------+ |
| | | |
| | | +---------------------+ +----------------+ |
| | | | Message Processing | | Security | |
| | | | Subsystem | | Subsystem | |
| | | | +------------+ | | | |
| | | | +->| v1MP * |<--->| +------------+ | |
| | | | | +------------+ | | | Other | | |
| | | | | +------------+ | | | Security | | |
| | | | +->| v2cMP * |<--->| | Model | | |
| | Message | | | +------------+ | | +------------+ | |
| | Dispatcher <--------->| +------------+ | | +------------+ | |
| | | | +->| v3MP * |<--->| | User-based | | |
| | | | | +------------+ | | | Security | | |
| | PDU Dispatcher | | | +------------+ | | | Model | | |
| +-------------------+ | +->| otherMP * |<--->| +------------+ | |
| ^ | +------------+ | | | |
| | +---------------------+ +----------------+ |
| v |
| +-------+-------------------------+---------------+ |
| ^ ^ ^ |
| | | | |
| v v v |
| +-------------+ +---------+ +--------------+ +-------------+ |
| | COMMAND | | ACCESS | | NOTIFICATION | | PROXY | |
| | RESPONDER |<->| CONTROL |<->| ORIGINATOR | | FORWARDER | |
| | application | | | | applications | | application | |
| +-------------+ +---------+ +--------------+ +-------------+ |
| ^ ^ |
| | | |
| v v |
| +----------------------------------------------+ |
| | MIB instrumentation | SNMP entity |
+-------------------------------------------------------------------+

3.2.1 USM and the diagram taken from RFC3411 Architecture

This following diagrams illustrate the difference in the security
processing done by the USM model and the security processing done by
a TMSM model.

The USM security model is encapsulated by the messaging model,
because the messaging model needs to (for incoming messages) 1)
decode the ASN.1 (messaging model) 2) determine the SNMP
architecture document.

TODO: Insert drawing here...

3.3 Passing messages between Dispatchers

Typically, security
model and parameters (messaging model) 3) decrypt the encrypted
portions of the message (security model) 4) translate parameters to
model-independent parameters (security model) 5) determine which
application should get the decrypted portions (messaging model), and
6) pass on the decrypted portions with model-independent parameters.

The USM approach uses SNMP-specific message security and parameters.

| -----------------------------------------------|
| transport layer |
| -----------------------------------------------|
^
|
v
--------------------------------------------------
| -----------------------------------------------|
| | transport mapping |
| -----------------------------------------------|
| ^
| |
| v
| --------------------------------------------- |
| --------------------- ------------------ |
| SNMP messaging <--> | decryption + | |
| | translation | |
| --------------------- ------------------ |
| ^
| |
| v
| --------------------- ------------------ |
| | SNMP applications | <--> | access control | |
| --------------------- ------------------ |

| --------------------------------------------- |

3.2.2 TMSM and the RFC3411 Architecture

In the TMSM approach, the order of the steps differ and may be
handled by different subsystems: 1) decrypt the encrypted portions of
the message (transport layer) 2) determine the SNMP security model
and parameters (transport mapping) 3*) translate parameters to model-
independent parameters (transport mapping) 4) decode the ASN.1
(messaging model) 5) determine which application should get the
decrypted portions (messaging model) 6*) translate parameters to
model-independent parameters (security model) 7) pass on the
decrypted portions with model-independent security parameters This is
largely based on having non-SNMP-specific message security and
parameters. The transport mapping model might provide the
translation from (e.g.) TLS user to securityName in step 3, OR The
TLS user might be passed to the messaging model to pass to a TMSM model,
security model to do the translation in step 6, if the WG decides all
translations should use the same translation table (e.g. the USM
MIB).

| -----------------------------------------------|
| ------------------ |
| transport layer <--> | decryption | |
| ------------------ |
| -----------------------------------------------|
^
|
v
--------------------------------------------------
| -----------------------------------------------|
| ------------------ |
| transport mapping <--> | translation* | |
| ------------------ |
| -----------------------------------------------|
| ^
| |
| v
| --------------------------------------------- |
| ------------------ |
| SNMP messaging <--> | translation* | |
| ------------------ |
| --------------------- ------------------ |
| ^
| |
| v
| --------------------- ------------------ |
| | SNMP applications | <--> | access control | |
| --------------------- ------------------ |

| --------------------------------------------- |

3.3 Passing Messages between Subsystems

RFC3411 defines ASIs that describe the passing of messages between
subsystem within an engine, and the parameters which are expected to
be passed between the subsystems. The ASIs generally pass model-
independent information.

A TMSM model will establish an encrypted tunnel between the transport
mappings of two SNMP engines,
without passing anything to the SNMP dispatcher. engines. One transport mapping security model
instance encrypts all messages, and the other transport mapping
security model instance decrypts the messages.

After the transport layer tunnel is established, then SNMP messages
can conceptually be sent through the tunnel from one SNMP engine message
dispatcher to another SNMP engine dispatcher. SNMP messages are
passed unencrypted from the source dispatcher to its own TMSM, and
presented unencrypted to the destination SNMP message dispatcher. Once the tunnel is
established, multiple SNMP messages may be able to be passed through
the same tunnel.

Within an engine, outgoing SNMP messages are passed unencrypted from
the message dispatcher to the transport mapping, and incoming
messages are passed unencrypted from the transport mapping to the
message dispatcher.

3.4 Security Parameter Passing Requirement

[RFC3411] section 4 describes primitives to describe the abstract
service interfaces used to conceptually pass information between the
various subsystems, models and applications within the architecture.
A Transport mapping Security Model

The security parameters include a model-independent identifier of the
security "principal", the security model used to perform the
authentication, and which SNMP-specific security services were
(should be) applied to the message (authentication and/or privacy).

In the RFC3411 architecture, the messaging model must pass information between
subsystems unpack SNMP-
specific security parameters from the message before calling a
security model to authenticate and decrypt an incoming message,
perform integrity checking, and translate model-specific security
parameters into model-independent parameters. In the TMSM approach,
the security -model specific parameters are not all carried in the
SNMP message, and can be determined from the transport layer by the
transport mapping, before the message processing begins.

[todo] For outgoing messages, it is necessary to have an MPSP because
it is the MPSP that actually creates the message from it scomponent
parts. Does the MPSP need to know the transport address or the
actual transport security capabilities, or can this be handled in the
TMSP, given the model-independent (and message-version-independent)
parameters? Are there any security services provided by the MPSP for
an outgoing message?

[todo] For incoming messages, is there security functionality that
can only be handled after the message version is known, such as well. the
comparison of transport security capabilities and msgFlags? Does
that functionality need to know the transport address and session or
just the model-independent security parameters (securityName, model,
level)? Are there any SNMP-specific parameters that need to be
unpacked from the message for MPSP handling? msgFlags, securityLevel,
etc.?

The RFC3411 architecture has no ASI parameters for passing security
information between the transport mapping and the dispatcher, and
between the dispatcher and the message processing model. Since the
TM portion of If there is
a need to have an MPSP called from the security message processing model to,
for example, verify that msgFlags and the MP portion of the transport security
model are co-resident within an implementation,
consistent, then it is assumed there
is a trust relationship that exists within will be necessary to pass the model-independent
security parameters from the TPSP through to the implementation. MPSP.

There are four approaches that could be used for passing information
between the TM portion of the securitymodel TMSP and the MP portion of the
security model : an MPSP.
1. we could define an ASI to supplement the existing ASIs, or
the TMSM could pass the information in an implementation-specific
cache, or
2. the TMSM could add a header to encapsulate the SNMP message, or
3. the TMSM could utilize fields already defined in the existing
SNMPv3 message. message, or
4. the TMSM could pass the information in an implementation-specific
cache or via a MIB module.

3.4.1 Using Define an ASI

RFC3411 discusses Abstract Service Iinterface

Abstract service interfaces [RFC3411] are defined by a set of
primitives that specify the purpose, services provided and an explicit non-purpose, of the
ASI approach: "This modularity of specification is not meant abstract data
elements that are to be
interpreted as imposing any specific requirements on implementation."
An ASI is not an API, and following a defined ASI is not required for
interoperability, so implementors passed when the services are really free invoked.
Defining additional ASIs to use any method
they choose. However, defining an ASI pass the security and transport
information from the transport mapping to a messaging security model
has the advantage of being consistent with existing RFC3411/3412 practice.

3.4.2 Using a cache

A cache mechanism could be used, into which the TM portion of the
security model puts information about the security applied to an
incoming message,
practice, and an MP portion of the security model extracts helps to ensure that information from any TMSM proposals pass the cache. The cache is
necessary data, and do not passed via an
explicit ASI. Given cause side effects by creating model-
specific dependencies between itself and other models or other
subsystems other than those that there may be multiple TM-security caches, a
cache ID probably needs to be passed in the message in the ASI so the
MP portion of the security model knows which cache to consult. This
approach would be consistent with the securityStateReference cache
already being passed around in the ASI.

The cache could be thought of as are clearly defined by an additional parameter in the ASI.
The ASI would not need to be changed since the SNMPv3 WG expected
that additional parameters could be passed for value-add features of
specific implementations.

3.4.3

3.4.2 Using an encapsulating header

A header could encapsulate the SNMP message to pass necessary
information from the TM portion of the security model TMSP to the dispatcher and then to the MP portion of the a messaging
security model. The message header would be included in the
wholeMessage ASI parameter, and would be removed by a corresponding
messaging model. This would imply the (one and only) messaging
dispatcher would need to be modified to determine which SNMP message
version was involved, and a new messaging message processing model would need
to be specified as well. The
other approaches may be able developed that knew how to use extract the header from the standard SNMPv3 messaging
model, with a new MP-security model.

3.4.4 Using existing fields in a message
and pass it to the MPSP.

3.4.3 Modifying Existing Fields in an SNMP Message

[RFC3412] describes the SNMPv3 message, which contains fields to pass
security related parameters. The TMSM could use these fields in an
SNMPv3 message, or comparable fields in other message formats to pass
information between transport mapping security models in different
SNMP engines, and to pass information between a TM transport mapping
security model and
the a corresponding MP messaging security model.

It is importnat to understand that SNMP messages are ASN.1 encoded,
and the SNMP architecture places no constraints on how the ASN.1 gets
decoded - it might be decoded in one massive decode, or individual
portions of the message, such as individual varbinds, may be decoded
only as needed. This is an implementation decision.

If the fields in an incoming SNMPv3 message are changed by the TM
portion TMSP
before passing it to the MP portion, MPSP, then the TM portion TMSP will need to encode its parameters decode the
ASN.1 message, modify the fields, and re-encode the message in ASN.1 or
before passing the message model would
need to be modified to permit non-encoded data on to be added the message dispatcher or to the
transport layer. This would require an intimate knowledge of the
message in a manner that format and message versions so the TMSP knew which fields
could be modified. This would not impact seriously violate the existing ASN.1
encoding/decoding modularity of
the message. In addition, the MP portion may
not architecture.

3.4.4 Using a cache

A cache mechanism could be able used, into which the TMSP puts information
about the security applied to perform a transport-independent message integrity
check, an incoming message, and transport-independent encryption an MPSP
extracts that information from the cache. Given that there may not be able
multiple TM-security caches, a cache ID would need to be
performed by passed
through an ASI so the MP portion MPSP knows which cache of the model. While it may be desirable
for most TMSM models information to perform those services through the TM portion
consult.

The cache reference could be thought of as an additional parameter in
the model, assuming ASIs between the use transport mapping and the messaging security
model. The RFC3411 ASIs would not need to be changed since the
SNMPv3 WG expected that additional parameters could be passed for
value-add features of specific implementations.

This approach does create dependencies between a cache or an encapsulating header model-specific TPSP
and a corresponding specific MPSP. If a TMSM-model-independent ASI
parameter is passed, this approach would not impose such constraints on future models. be consistent with the
securityStateReference cache already being passed around in the ASI.

This document will describe a cache approac
h, but an encapsulating
header or other mechanisms could also be used if preferred for
specific TM security models. cache-based approach.

3.5 Architectural Requirements for Access Control Requirements

3.5.1 Architectural securityName Binding Requirement

For SNMP access control to function properly, the security mechanism
must establish a security model identifier, a securityLevel, and a
securityName, which is the security model independent identifier for
a principal, a security model identifier,
and a securityLevel. principal. The SNMPv3 message processing architecture subsystem
relies on a message model based security model, such as USM, to play an a role in security
that goes beyond protecting the message - it ties various security models for provides a mapping
between the same USM-specific principal to a security-model independent
securityName which can be used for subsequent processing, such as for
access control.

The TMSM assumes two portions to is a two-stage security model, one tied to the with a transport mapping
security process (TMSP) and another tied to the a message processing model.
and will be referred to here as a TM-portion and an MP-portion of the security model. process
(MPSP). Depending on the specific design of the security specific TMSM model, i.e.

which transport layer protocol is used, different features might be
provided by the TM portion TMSP or by the MP portion. MPSP. For example, the binding of translation
from a mechanism-specific authenticated identity to a securityName
might be done by the TM
portion TMSP or by the MP portion. MPSP. [todo] It may be possible
to define a consistent division of stages regardless of the transport
layer protocol used, and a consistent division of functionality would
be preferred.

The SNMP architecture distinguishes between messages with no
authentication and no privacy (noAuthNoPriv), authentication without
privacy (authNoPriv) and authentication with privacy (authPriv).
Hence, the authentication of a transport layer identity plays an
important role and must be considered by any TMSM, and user
authentication must be available via the transport layer security
mechanism used. However, it
protocol.

If the type of authentication provided by the transport layer (e.g.
host-based or anonymous) is also possible that considered adequate to secure and/or
encrypt the message, but inadequate to provide the desired
granularity of access control (e.g. user-based), a second level of
authentication, e.g. one provided by a AAA server, for example, may be used to
provide the authentication identity which is bound to the
securityName, if
securityName. This approach would require a good analysis of the type
potential for man-in-the-middle attacks or masquerade possibilities.

3.5.2 Separation of Authentication and Authorization

A TMSM security model should take care to not violate the separation
of authentication provided by and authorization in the transport
layer (e.g. host-based or anonymous) RFC3411 architecture..
The isAccessAllowed() primitive is considered adequate used for passing security-model
independent parameters between the subsystems of the architecture.

Mapping of (securityModel, securityName) to
secure and/or encrypt an access control policy
should be handled within the message, but inadequate access control subsystem, not the
security subsystem, to be consistent with the modularity of the
RFC3411 architecture. This separation was a deliberate decision of
the SNMPv3 WG, to allow support for authentication protocols which
did not provide authorization capabilities, and to support
authorization schemes, such as VACM, that do not perform their own
authentication.

An authorization model MAY require authentication by certain
securityModels and a minimum securityLevel to allow access to the
desired granularity
data.

TMSM is an enhancement for the SNMPv3 privacy and authentication
provisions, but it is not a significant improvement for the
authorization needs of SNMPv3. TMSM provides all the model-
independent parameters for the isAccessAllowed() primitive [RFC3411].

TMSM does not specify how the securityModel and securityName could be
dynamically mapped to a VACM-style groupName. The mapping of
(securityModel, securityName) to a groupName is a VACM-specific
mechanism for naming an access control policy, and for tying the
named policy to the addressing capabilities of the data modeling
language (e.g. user-based). SMIv2), the operations supported, and other factors.
Providing a binding outside the Access Control subsystem might create
dependencies that could make it harder to develop alternate models of
access control, such as one built on UNIX groups, Windows domains,
XML hierarchies, or task-based controls. The preferred approach is
to pass the model-independent security parameters via the
isAccessAllowed() ASI, and perform the mapping within the access
control model.

To provide support for protocols which simultaneously send
information for authentication and authorization, such as RADIUS,
model-specific authorization information MAY be cached or otherwise
made available to the access control subsystem, e.g. via a MIB table
similar to the vacmSecurityToGroupTable, so the access control
subsystem can create an approrpiate binding between the model-
independent securityModel and securityName and a model-specific
access control policy. This may be highly undesirable, however, if
it creates a dependency between a security model and an access
control model, just as it is undesirable that the TMSM approach
creates a dependency between a TMSP and an MPSP.

4. Fields in Integration with the SNMPv3 message format

TMSM proposals can use the SNMPv3 message format, defined in RFC3412,
section 6. This seection discusses how the fields could be reused.

4.1 msgVersion

For proposals that reuse the SNMPv3 message format, this field should
contain the value 3.

4.2 msgGlobalData

msgID and msgMaxSize are used identically for the TMSM models as for
the USM model.

msgSecurityModel should be set to a value from the SnmpSecurityModel
enumeration [RFC3410] [RFC3411] to identify the specific TMSM model. Each
standards-track TMSM model should have an enumeration assigned by
IANA. Each enterprise-specific security model should have an
enumeration assigned following instructions in the description of the
SnmpSecurityModel TEXTUAL-CONVENTION from RFC3411.

msgSecurityParameters would carry security information required for
message security processing. It is unclear whether this field would
be useful or what parameters would be carried to support security,
since message security is provided by an external process, and
msgSecurityParameters are not used identically for by the access control subsystem.

RFC3412 defines two primitives, generateRequestMsg() and
processIncomingMsg() which require the specification of an
authoritative SNMP entity. [todo] We need to discuss what the meaning
of authoritative would be in a TMSM models as for environment, whether the specific
services provided in USM model. security from msgSecurityParameters still
are needed, and how the Message Processing model provides this
information to the security model via generateRequestMsg() and
processIncomingMsg() primitives. RFC3412 specifies that "The data in
the msgSecurityParameters field is used exclusively by the Security
Model, and the contents and format of the data is defined by the
Security Model. This OCTET STRING is not interpreted by the v3MP,
but is passed to the local implementation of the Security Model
indicated by the msgSecurityModel field in the message."

msgFlags have the same values for the TMSM models as for the USM
model. "The authFlag and privFlag fields indicate the securityLevel
that was applied to the message before it was sent on the wire."

4.3 securityLevel and msgFlags

For an outgoing message, msgFlags is the requested security for the
message; if a TMSM cannot provide the requested securityLevel, the
model MUST describe a standard behavior that is followed for that
situation. If the TMSM cannot provide at least the requested level
of security, the TMSM MUST discard the request and SHOULD notify the
message processing model that the request failed. [dbh: how is yet to
be determined, and may be model-specific or implementation-specific.]

For an outgoing message, if the TMSM is able to provide stronger
than requested security, that may be acceptable. The transport layer
protocol would need to indicate to the receiver what security has
been applied to the actual message. To avoid the need to mess with
the ASN.1 encoding, the SNMPv3 message carries the requested
msgFlags, not the actual securityLevel applied to the message. If a
message format other than SNMPv3 is used, then the new message may
carry the more accurate securityLevel in the SNMP message.

For an incoming message, the receiving TMSM knows what must be done
to process the message based on the transport layer mechanisms. If
the underlying transport security mechanisms for the receiver cannot
provide the matching securityLevel, then the message should follow
the standard behaviors for the transport security mechanism, or be
discarded silently.

Part of the responsibility of the TMSM is to ensure that the actual
security provided by the underlying transport layer security
mechanisms is configured to meet or exceed the securityLevel required
by the msgFlags in the SNMP message. When the MP portion of the
security model MPSP processes the
incoming message, it should compare the msgFlags field to the
securityLevel actually provided for the message by the transport
layer security. If they differ, the MP portion of
the security model MPSP should determine whether
the changed securityLevel is acceptable. If not, it should discard
the message. Depending on the model, the MP portion MPSP may issue a reportPDU
with the XXXXXXX model-specific counter.

Questions about msgFlags:

Is the securityLevel looked at before the security model gets to
it.? No. the security model has two parts - the TM portion and
the MP portion. The securityLevel is looked at by the TM portion
before it gets to the MP piece, but both are parts of the same
security model.
Would it be legal for the security model to ignore the incoming
flags and change them before passing them back up? If it changed
them, it wouldn't necessarily be ignoring them. The TM portion
should pass both an actual securityLevel applied to the message,
and the msgFlags in the SNMP message to the MP piece for
consideration related to access control.. The msgFlags parameter
in the SNMP message is never changed when processing an incoming
message.
Would it be legal for the security model to ignore the outgoing
flags and change them before passing them out? no; because the two
portions are parts of the same security model, either the MP piece
should recognize that a securityLevel cannot be met or exceeded,
and reject the message during the message-build phase, or the TM
piece should determine if it is possible to honor the request. It
is possible to apply an increased securityLevel for an outgoing
request, but the procedure to do so must be spelled out clearly in
the model design.
The security model would need to (MUST) check the incoming
security level flags to make sure they matched the TLS/whatever
session setup and if not drop the message. Yes, mostly.
Depending on the model, either the TM portion or the MP portion
MUST verify that the actual processing met or exceeded the
securityLevel requested by the msgFlags and that it is acceptable
to the specific-model processing (or operator configuration) for
this different securityLevel to be applied to the message. This
is also true (especially) for outgoing messages.
You might legally be able to have a authNoPriv message that is
actually encrypted via the transport (but not the other way around
of course). Yes, a TMSM could define that as the behavior (or
permit an operator to specify that is acceptable behavior) when a
requested securityLevel cannot be provided, but a stronger
securityLevel can be provided.

See the Appendix A appendix for further discussion of the msgFlags
field ve
rsus the actual securityLevel provided. [dbh: it may be a
good thing to merge the Question and Answer with the appendix, either
here or there.]

4.4 The tmStateReference for Passing Security Parameters

A tmStateReference is used to pass data between the TM portion TMSP and the MP portion of the security model,
MPSP, similar to the securityStateReference described in RFC3412.
This can be envisioned as being appended to the ASIs between the TM
and the MP or as being passed in an encapsulating header.

The TM portion of the security model TMSP may provide only some aspects of security, and leave some
aspects to the MP portion of the model. MPSP. tmStateReference should be used to pass any
parameters, in a model- and mechanism-specific format, that will be
needed to coordinate the activities of the TM TMSP and MP portions of the model, MPSP, and the
parameters subsequently passed in securityStateReference . For
example, the TM
portion TMSP may provide privacy and data integrity and
authentication and authorization policy retrievals, or some subset of
these features, depending on the features available in the transport
mechanisms. A field in tmStateReference should identify which
services were provided for each received message by the TM portion, TMSP, the
securityLevel applied to the received message, the model-specific
security identity, the session identifier for session based transport
security, and so on.

4.5 securityStateReference Cached Security Data

From RFC3411: "For each message received, the Security Model caches
the state information such that a Response message can be generated
using the same security information, even if the Local Configuration
Datastore is altered between the time of the incoming request and the
outgoing response.

A Message Processing Model has the responsibility for explicitly
releasing the cached data if such data is no longer needed. To
enable this, an abstract securityStateReference data element is
passed from the Security Model to the Message Processing Model. The
cached security data may be implicitly released via the generation of
a response, or explicitly released by using the stateRelease
primitive, as described in RFC3411 section 4.5.1."

For the TMSM approach, the TMSP may need to provide information to
the message processing model, such as the security-model-independent
securityName, securityLevel, and securityModel parameters, and for
responses, the messaging model may need to pass the parameters back
to the TMSP. To differentiate what information needs to be provided
to the MP
portion message processing model by the TM portion, TMSP, and vice-versa, this
document will differentiate the tmStateReference provide by the TMSP
from the securityStateReference. securityStateReference provided by the MPSP. An
implementation MAY use one cache and one reference to serve both
functions, but an implementor must be aware of the cache-release
issues to prevent the cache from being released before the TM portion transport
mapping has had an opportunity to extract the information it needs.

4.5.1 Prepare an Outgoing SNMP Message

According to

Following RFC3412, section 7.1, the SNMPv3 message processing model calls the MP portion of the TM security model using
uses the generateResponseMsg() or generateRequestMsg(). generateRequestMsg() primitives, to
call the MPSP. The MP portion of message processing model, or the
model MPSP it calls,
may need to put information into the tmStateReference cache for use
by the TM portion of the model, TMSP, such as:
tmSecurityStateReference - the unique identifier for the cached
information
tmTransportDomain
tmTransportAddress
tmSecurityModel - an indicator of which mechanisms to use
tmSecurityName - a model-specific identifier of the security
principal
tmSecurityLevel - an indicator of which security services are
requested
and may contain additional information such as
tmSessionID
tmSessionKey
tmSessionMsgID

According to RFC3411, section 4.1.1, the application provides the
transportDomain and transportAddress to the PDU dispatcher via the
sendPDU() primitive. If we permit multiple sessions per
transportAddress, then we would need to define how session
identifiers get passed from the application to the PDU dispatcher
(and then to the MP model).

The SNMP over TCP Transport Mapping document [RFC3430] says that TCP
connections can be recreated dynamically or kept for future use and
actually leaves all that to the transport mapping.

[todo] we might define a new transportDomain and transportAddress,
which includes the address and session identifier. For situations
where a session has not yet been established, we could pass a 0x0000
session identifier (or whatever) to indicate that a session should be
established.

We might have a MIB module that records the session information for
subsequent use by the applications and other subsytems, or it might
be passed in the tmStateReference cache. For notifications, I assume
the SNMPv3 notification tables would be a place to find the address,
but I'm not sure how to identify the presumably-dynamic session
identifiers. The MIB module could identify whether the session was
initiated by the remote engine or initiated by the current engine,
and possibly assigned a purpose (incoming request/response or
outgoing notifications). First we need to decide whether to handle
notifications and requests in one or two (or more) sessions, which
might depend on the transport protocol we choose (the same problem
netconf faced).

4.5.2 Prepare Data Elements from an Incoming SNMP Message

For an incoming message, the TM portion of a model TMSP will need to put information from
the transport mechanisms used into the tmStateReference so the MP portion of the model MPSP
can extract the information and add it conceptually to the
securityStateReference.

The tmStateReference cache will likely contain at least the following
information:
tmStateReference - a unique identifier for the cached information
tmSecurityStateReference - the unique identifier for the cached
information
tmTransportDomain
tmTransportAddress
tmSecurityModel - an indicator of which mechanisms to use
tmSecurityName - a model-specific identifier of the security
principal
tmSecurityLevel - an indicator of which security services are
requested
tmAuthProtocol
tmPrivProtocol
and may contain additional information such as
tmSessionID
tmSessionKey
tmSessionMsgID

4.6 Notifications

For notifications, if the cache has been released and then session
closed, then the MP portion of the security model MPSP will request the TM
portion of the security model TMSP to establish a session,
populate the cache, and pass the securityStateReference to the MP portion of the
security model.

TODO: MPSP.

[todo] We need to determine what state needs to be saved here.

5. Transport Mapping Security Model Samples

Using a protocol in a manner for which is was not designed has
numerous problems. The advertised security characteristics of a
protocol may depend on its being used as designed; when used in other
ways, it may not deliver the expected security characteristics. It
is recommended that any proposed model include a discussion of the
applicability statement of the protocols to be used.

5.1 TLS/TCP Transport Mapping Security Model

SNMP supports multiple transports. The preferred transport for SNMP
over IP is UDP [RFC3417]. An experimental transport for SNMP over
TCP is defined in [RFC3430].

TLS/TCP will create an association between the TMSM of one SNMP
entity and the TMSM of another SNMP entity. The created "tunnel" may
provide encryption and data integrity. Both encryption and data
integrity are optional features in TLS. The TLS TM portion of the
security model TMSP MUST provide
authentication if auth is requested in the securityLevel of the SNMP
message request (RFC3412 4.1.1). The TLS TM-security model MUST
specify that the messages be encrypted if priv is requested in the
securityLevel parameter of the SNMP message request (RFC3412 4.1.1).

The TLS TM-security model SHOULD use MUST support the TLS Handshake Protocol
with mutual authentication.

5.1.1 tmStateReference for TLS

Upon establishment of a TLS session, the TM-security model TMSP will cache the state
information. A tmStateReference that is unique within the
SNMP entity tmStateReference will be stored in the cache, and passed to the
corresponding MP portion of the security model, to enable lookup. MPSP. The MP security model MPSP will pass the securityStateReference to
the Message Processing Model for memory management.

The tmStateReference cache:
tmStateReference
tmSecurityStateReference
tmTransportDomain = TCP/IPv4
tmTransportAddress = x.x.x.x:y
tmSecurityModel - TLS TMSM
tmSecurityName = "dbharrington"
tmSecurityLevel = "authPriv"
tmAuthProtocol = Handshake MD5
tmPrivProtocol = Handshake DES
tmSessionID = Handshake session identifier
tmSessionKey = Handshake peer certificate
tmSessionMasterSecret = master secret
tmSessionParameters = compression method, cipher spec,
is-resumable is-
resumable

5.1.2 MP portion MPSP for TLS TM-Security Model

messageProcessingModel = SNMPv3
securityModel = TLS TMSM
securityName = tmSecurityName
securityLevel = msgSecurityLevel

5.1.3 MIB Module for TLS Security

Each security model should use its own MIB module, rather than
utilizing the USM MIB, to eliminate dependencies on a model that
could be replaced some day. See RFC3411 section 4.1.1.

The TLS MIB module needs to provide the mapping from model-specific
identity to a model independent securityName.

TO
DO:

[todo] Module needs to be worked out once things become stable...

5.2 DTLS/UDP Transport Mapping Security Model

DTLS has been proposed as a UDP-based TLS. Transport Layer Security
(TLS) [RFC2246] traditionally requires a connection-oriented
transport and is usually used over TCP. Datagram Transport Layer
Security (DTLS) [DTLS] provides security services equivalent to TLS
for connection-less transports such as UDP.

DTLS provides all the security services needed from an SNMP
architectural point of view. Although it is possible to derive a
securityName from the public key certificates (e.g. the subject
field), this approach requires to install installing certificates on agents and
as well as managers, all SNMP
entities, leading to a certificate management problem which again does not
integrate well with established AAA systems. [todo] why does this not
integrate well with existing AAA systems?
Another option is to run an authentication exchange which is
integrated with TLS, such as Secure Remote Password with TLS

[SRP-TLS]. [SRP-
TLS]. A similar option would be to use Kerberos authentication with
TLS as defined in [RFC2712].

It is important to stress that the authentication exchange must be
integrated into the TLS mechanism to prevent man-in-the-middle
attacks. While SASL [RFC2222] is often used on top of a TLS
encrypted channel to authenticate users, this choice seems to be
problematic until the mechanism to cryptographically bind SASL into
the TLS mechanism has been defined.

DTLS will create an association between the TMSM of one SNMP entity
and the TMSM of another SNMP entity. The created "tunnel" may
provide encryption and data integrity. Both encryption and data
integrity are optional features in DTLS. The DTLS TM-security model
MUST provide authentication if auth is requested in the securityLevel
of the SNMP message request (RFC3412 4.1.1). The TLS TM-security
model MUST specify that the messages be encrypted if priv is
requested in the securityLevel parameter of the SNMP message request
(RFC3412 4.1.1).

The DTLS TM-security model SHOULD use MUST support the TLS Handshake Protocol
with mutual authentication.

5.2.1 tmStateReference for DTLS

Upon establishment of a DTLS session, the TM-security model TMSP will cache the state
information. A tmStateReference that is unique
within the SNMP entity tmStateReference will be stored in the cache, and passed to the
corresponding MP portion of the security model, to enable lookup. MPSP. The MP security model MPSP will pass the securityStateReference to
the Message Processing Model for memory management.

The tmStateReference cache:
tmStateReference
tmSecurityStateReference
tmTransportDomain = UDP/IPv4
tmTransportAddress = x.x.x.x:y
tmSecurityModel - DTLS TMSM
tmSecurityName = "dbharrington"
tmSecurityLevel = "authPriv"
tmAuthProtocol = Handshake MD5
tmPrivProtocol = Handshake DES
tmSessionID = Handshake session identifier
tmSessionKey = Handshake peer certificate
tmSessionMasterSecret = master secret
tmSessionParameters = compression method, cipher spec,
is-resumable is-
resumable
tmSessionSequence = epoch, sequence

TODO:

[todo]
Need to discuss to what extent DTLS is a reasonable choice for
SNMP interactions.
What is the status of the work to cryptographically bind SASL to
DTLS?
More details need to be worked out...

5.3 SASL Transport Mapping Security Model

The Simple Authentication and Security Layer (SASL) [RFC2222]
provides a hook for authentication and security mechanisms to be used
in application protocols. SASL supports a number of authentication
and security mechanisms, among them Kerberos via the GSSAPI
mechanism.

This sample will use DIGEST-MD5 because it supports authentication,
integrity checking, and confidentiality.

DIGEST-MD5 supports auth, auth with integrity, and auth with
confidentiality. Since SNMPv3 assumes integrity checking is part of
authentication, if msgFlags is set to authNoPriv, the qop-value
should be set to auth-int; if msgFlags is authPriv, then qop-value
should be auth-conf.

Realm is optional, but can be utilized by the securityModel if
desired. SNMP does not use this value, but a TMSM could map the
realm into SNMP processing in various ways. For example, realm and
username could be concatenated to be the securityName value, e.g.
helpdesk::username", or the realm could be used to specify a
groupname to use in the VACM access control. This would be similar
to the EUSM's approach to having the securityName-to-group mapping done by the external AAA
server.

5.3.1 tmStateReference for SASL DIGEST-MD5

The tmStateReference cache:
tmStateReference
tmSecurityStateReference
tmTransportDomain = TCP/IPv4
tmTransportAddress = x.x.x.x:y
tmSecurityModel - SASL TMSM
tmSecurityName = username
tmSecurityLevel = [auth-conf]
tmAuthProtocol = md5-sess
tmPrivProtocol = 3des
tmServicesProvided =
mutual authentication,
reauthentication,
integrity,
encryption
tmParameters = "realm=helpdesk, serv-type=SNMP

6. Security Considerations

This document describes an architectural approach and multiple
proposed configurations that would permit SNMPv3 to utilize transport
layer security services. Each section containing a proposal should
discuss the security considerations of that approach. [todo] expand
as needed.

Perfect forward secrecy guarantees that compromise of long term
secret keys does not result in disclosure of past session keys.

It is considered desirable by some industry segements that TMSM
security models should utilize transport layer security that
addresses perfect forward secrecy at least for encryption keys.

7. Acknowledgments

The authors would like to thank Ira McDonald, Ken Hornstein, and
Nagendra Modadugu for their comments and suggestions.

8. References

7.1

8.1 Normative References

[RFC3411] Harrington, D., Presuhn, R. R., and B. Wijnen, "An
Architecture for Describing Simple Network Management
Protocol (SNMP) Management Frameworks", STD 62, RFC 3411,
December 2002.

[RFC3412] Case, J., Harrington, D., Presuhn, R. R., and B. Wijnen,
"Message processing and Dispatching for SNMP", STD 62,
RFC 3412, December 2002.

[RFC3414] Blumenthal, U. and B. Wijnen, "User-based Security Model
(USM) for version 3 of the Simple Network Management
Protocol (SNMPv3)", STD 62, RFC 3414, December 2002.

[RFC3417] Presuhn (Editor), R., "Transport Mappings for the Simple
Network Management Protocol (SNMP)", STD 62, RFC 3417,
December 2002.

[RFC3430] Schoenwaelder, J., "Simple Network Management Protocol
(SNMP) over Transmission Control Protocol (TCP) Transport
Mapping", RFC 3430, December 2002.

[RFC2246] Dierks, T. and C. Allen, "The TLS Protocol Version 1.0",
RFC 2246, January 1999.

[RFC1510] Kohl, J. and B. Neuman, "The Kerberos Network
Authentication Service (V5)", RFC 1510, September 1993.

[RFC2222] Myers, J., "Simple Authentication and Security Layer
(SASL)", STD 62, RFC RFC2222, October 1997.

[DTLS] Rescorla, E. and N. Modadugu, "Datagram Transport Layer
Security", ID draft-rescorla-dtls-01.txt, July 2004.

7.2

8.2 Informative References

[RFC3410] Case, J., Mundy, R., Partain, D. D., and B. Stewart,
"Introduction and Applicability Statements for
Internet-Standard Internet-
Standard Management Framework", RFC 3410, December 2002.

[RFC2712] Medvinsky, A. and M. Hur, "Addition of Kerberos Cipher
Suites to Transport Layer Security (TLS)", RFC 2712,
October 1999.

[SRP-TLS] Taylor, D., Wu, T., Mavroyanopoulos, M. M., and T. Perrin,
"Using SRP for TLS Authentication",
ID draft-ietf-tls-srp-08.txt, August 2004.

[EUSM] Narayan, D., McCloghrie, K., Salowey, J. J., and C. Elliot,
"External USM for SNMPv3",
ID draft-kaushik-snmp-external-usm-00.txt, July 2004.

[NETCONF] Enns, R., "NETCONF Configuration Protocol",
ID draft-ietf-netconf-prot-04.txt, October 2004.

[SSHauth] Lonvick, C., "SSH Authentication Protocol",
ID draft-ietf-secsh-userauth-21.txt, June 2004.

Authors' Addresses

David Harrington
Independent
Harding Rd
Portsmouth NH
USA

Phone: +1 603 436 8634
EMail:
Email: dbharrington@comcast.net

Juergen Schoenwaelder
International University Bremen
Campus Ring 1
28725 Bremen
Germany

Phone: +49 421 200-3587
EMail:
Email: j.schoenwaelder@iu-bremen.de

Appendix A. Message Questions about msgFlags:

[todo] many of these questions can be resolved by deciding whether
the TMSP or MPSP provides the service of comparing msgFlags (from
inside the message) to actual capabilities of the transport layer
security versus (external to the message). It may however be necessary to
provide this service for two slightly different purposes depending on
whether the message is outgoing (and may need to be checked by the
TMSP when a new transport session might be created) or the message is
incoming ( the capabilities of the transport layer session are
already known, but msgFlags has not been unpacked yet at the TMSP, so
the comparison must be done at the MPSP). Of course, we really only
need to compare the authflag and the privflag, i.e. the
securityLevel, so if we pass the securityLevel between the two
stages, then they each have the info they need to do their respective
comparisons.

There have been a large number of questions about msgFlags in the
TMSM approach, mostly concerning the msgFlags value and the actual
security provided, and whether msgFlags can be used to initiate per-
message or per-session security.

A.1 msgFlags versus actual security

Using IPSEC, SSH, or SSL/TLS to provide security services "below" the
SNMP message, the use of securityName and securityLevel will differ
from the USM/VACM approach to SNMP access control. VACM uses the
"securityName" and the "securityLevel" to determine if access is
allowed. With the SNMPv3 message and USM security model, both
securityLevel and securityName are contained in every SNMPv3 message.

Any proposal for a security model using IPSEC, SSH, or SSL/TLS needs
to specify how this info is made available to the SNMPv3 message
processing, and how it is used.

One specific case to consider is the relationship between the
msgFlags of an SNMPv3 message, and the actual services provided by
the lower layer security. For example, if a session is set up with
encryption, is the priv bit always (or never) set in the msgFlags
field, and is the PDU never (or always) encrypted? Do msgFlags have
to match the security services provided by the lower layer, or are
the msgFlags ignored and the values from the lower layer used?

Is the securityLevel looked at before the security model gets to
it.? No. the security model has two parts - the TMSP and the
MPSP. The securityLevel is looked at by the TMSP before it gets
to the MPSP, but both are parts of the same security model.
Would it be legal for the security model to ignore the incoming
flags and change them before passing them back up? If it changed
them, it wouldn't necessarily be ignoring them. The TMSP should
pass both an actual securityLevel applied to the message, and the
msgFlags in the SNMP message to the MPSP for consideration related
to access control.. The msgFlags parameter in the SNMP message is
never changed when processing an incoming message.
Would it be legal for the security model to ignore the outgoing
flags and change them before passing them out? no; because the two
stages are parts of the same security model, either the MPSP
should recognize that a securityLevel cannot be met or exceeded,
and reject the message during the message-build phase, or the TMSP
should determine if it is possible to honor the request. It is
possible to apply an increased securityLevel for an outgoing
request, but the procedure to do so must be spelled out clearly in
the model design.
The security model MUST check the incoming security level flags to
make sure they matched the transport session setup. and if not
drop the message. Yes, mostly. Depending on the model, either
the TMSP or the MPSP MUST verify that the actual processing met or
exceeded the securityLevel requested by the msgFlags and that it
is acceptable to the specific-model processing (or operator
configuration) for this different securityLevel to be applied to
the message. This is also true (especially) for outgoing
messages.
You might legally be able to have a authNoPriv message that is
actually encrypted via the transport (but not the other way around
of course). Yes, a TMSM could define that as the behavior (or
permit an operator to specify that is acceptable behavior) when a
requested securityLevel cannot be provided, but a stronger
securityLevel can be provided.

A.2 Message security versus session security

For SBSM, and for many TMSM models, securityName is specified
during session setup, and associated with the session identifier.
Is it possible for the request (and notification) originator to
specify per message auth and encryption services, or are they are
"fixed" by the transport/session model?
If a session is created as 'authPriv', then keys for encryption
would still be negotiated once at the beginning of the session.
But if a message is presented to the session with a security level
of authNoPriv, then that message could simply be authenticated and
not encrypted. Wouldn't that also have some security benefit, in
that it reduces the encrypted data available to an attacker
gathering packets to try and discover the encryption keys?

Agents
Some SNMP entities are often resource-constrained. Adding sessions
increases the need for resources, we shouldn't require two
sessions when one can suffice. 2 bytes per session structure and a
compare or two is much less of a resource burden on an agent than two separate
sessions.
It's not just about CPU power of the device but the percentage of
CPU cycles that are spent on network management. There isn't much
value in using encryption for a performance management system
polling PEs for performance data on thousands of interfaces every
ten minutes,it just adds significant overhead to processing of
the packet. Using an encrypted TLS channel for everything may not
work for use cases in performance management wherein we collect
massive amounts of non sensitive data at periodic intervals. Each
SNMP "session" would have to negotiate two separate protection
channels (authPriv and authNoPriv) and for every packet the SNMP
engine will use the appropriate channel based on the desired
securityLevel.
If the underlying transport layer security was configurable on a
per-message basis, a TMSM could have a MIB module with
configurable maxSecurityLevel and a minSecurityLevel objects to
identify the range of possible levels, and not all messages sent
via that session are of the same level. A session's
maxSecurityLevel would identify the maximum security it could
provide, and a session created with a minSecurityLevel of authPriv
would reject an attempt to send an authNoPriv message.

Intellectual Property Statement

The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.

Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.

The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.

Disclaimer of Validity

This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM 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.

Copyright (C) The Internet Society (2004). (2005). This document is subject
to the rights, licenses and restrictions contained in BCP 78, and
except as set forth therein, the authors retain all their rights.

Acknowledgment

Funding for the RFC Editor function is currently provided by the
Internet Society.