Internet Engineering Task Force (IETF) G. Lebovitz
Request for Comments: 6862
Category: Informational M. Bhatia
ISSN: 2070-1721 Alcatel-Lucent
B. Weis
Cisco Systems
March 2013
Keying and Authentication for Routing Protocols (KARP)
Overview, Threats, and Requirements
Abstract
Different routing protocols employ different mechanisms for securing
protocol packets on the wire. While most already have some method
for accomplishing cryptographic message authentication, in many cases
the existing methods are dated, vulnerable to attack, and employ
cryptographic algorithms that have been deprecated. The "Keying and
Authentication for Routing Protocols" (KARP) effort aims to overhaul
and improve these mechanisms. This document does not contain
protocol specifications. Instead, it defines the areas where
protocol specification work is needed. This document is a companion
document to RFC 6518, "Keying and Authentication for Routing
Protocols (KARP) Design Guidelines"; together they form the guidance
and instruction KARP design teams will use to review and overhaul
routing protocol transport security.
Status of This Memo
This document is not an Internet Standards Track specification; it is
published for informational purposes.
This document is a product of the Internet Engineering Task Force
(IETF). It represents the consensus of the IETF community. It has
received public review and has been approved for publication by the
Internet Engineering Steering Group (IESG). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc6862.
Lebovitz, et al. Informational [Page 1]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
Copyright Notice
Copyright (c) 2013 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4
1.2. Requirements Language . . . . . . . . . . . . . . . . . . 7
2. KARP Effort Overview . . . . . . . . . . . . . . . . . . . . . 7
2.1. KARP Scope . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2. Incremental Approach . . . . . . . . . . . . . . . . . . . 8
2.3. Goals . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4. Non-Goals . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5. Audience . . . . . . . . . . . . . . . . . . . . . . . . . 12
3. Threats . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1. Threat Sources . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1. OUTSIDERS . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2. Unauthorized Key Holder . . . . . . . . . . . . . . . 14
3.1.2.1. Terminated Employee . . . . . . . . . . . . . . . 15
3.1.3. BYZANTINE . . . . . . . . . . . . . . . . . . . . . . 15
3.2. Threat Actions In Scope . . . . . . . . . . . . . . . . . 16
3.3. Threat Actions Out of Scope . . . . . . . . . . . . . . . 17
4. Requirements for KARP Work Phase 1: Update to a Routing
Protocol's Existing Transport Security . . . . . . . . . . . . 18
5. Security Considerations . . . . . . . . . . . . . . . . . . . 23
6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 24
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 24
7.1. Normative References . . . . . . . . . . . . . . . . . . . 24
7.2. Informative References . . . . . . . . . . . . . . . . . . 24
Lebovitz, et al. Informational [Page 2]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
1. Introduction
In March 2006, the Internet Architecture Board (IAB) held a workshop
on the topic "Unwanted Internet Traffic". The report from that
workshop is documented in [RFC4948]. Section 8.1 of that document
states, "A simple risk analysis would suggest that an ideal attack
target of minimal cost but maximal disruption is the core routing
infrastructure". Section 8.2 calls for "[t]ightening the security of
the core routing infrastructure". Four main steps were identified
for that tightening:
o Create secure mechanisms and practices for operating routers.
o Clean up the Internet Routing Registry (IRR) repository, and
secure both the database and the access to it, so that it can be
used for routing verification.
o Create specifications for cryptographic validation of routing
message content.
o Secure the routing protocols' packets on the wire
The first bullet is being addressed in the OPSEC working group. The
second bullet should be addressed through liaisons with those running
the IRR's globally. The third bullet is being addressed in other
efforts within the IETF. For example, BGP message content validity
is being addressed in the SIDR working group.
This document addresses the last item in the list above, securing the
transmission of routing protocol packets on the wire. More
precisely, it focuses on securing the transport systems employed by
routing protocols, including any mechanisms built into the protocols
themselves to authenticate packets. This effort is referred to as
Keying and Authentication for Routing Protocols, or "KARP". KARP is
concerned with issues and techniques for protecting the messages
between directly communicating peers. This type of protection may
overlap with, but is strongly distinct from, protection designed to
ensure that routing information is properly authorized relative to
the source of the information. Such assurances are provided by other
mechanisms and are outside the scope of this document.
This document is one of two that together form the guidance and
instructions for KARP design teams working to overhaul routing
protocol transport security. The other document is the KARP Design
Guide [RFC6518].
Lebovitz, et al. Informational [Page 3]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
This document does not contain protocol specifications. Instead, its
goal is to define the areas where protocol specification work is
needed and to provide a set of requirements for KARP design teams to
follow as they update a routing protocol's existing transport
security (see Work Phase 1 in Section 4.1 of [RFC6518]).
This document has three main parts. The first part, found in Section
2, provides an overview of the KARP effort. The second part, in
Section 3, lists the threats from "Generic Threats To Routing
Protocols" [RFC4593] that are in scope for per-packet authentication
for routing protocol transport systems. Therefore, this document
does not contain a complete threat model; it simply points to the
parts of the governing threat model that KARP design teams must
address and explicitly states which parts are out of scope for KARP
design teams. The third part, in Section 4, enumerates the
requirements that routing protocol specifications must meet when
addressing the threats related to KARP's Work Phase 1, the update to
a routing protocol's existing transport security. ("Work Phase 2", a
framework and usage of a Key Management Protocol (KMP), will be
addressed in a future document[s]).
1.1. Terminology
This document uses the terminology "on the wire" to refer to the
information used by routing protocols' transport systems. This term
is widely used in RFCs, but is used in several different ways. In
this document, it is used to refer both to information exchanged
between routing protocol instances and to underlying protocols that
may also need to be protected in specific circumstances. Individual
protocol analysis documents will need to be more specific in their
use of this phrase.
Additionally, within the scope of this document, the following words,
when beginning with a capital letter, or spelled in all capital
letters, hold the meanings described in this section. If the same
word is used uncapitalized, then it is intended to have its common
English definition.
Identifier
The type and value used by a peer of an authenticated message
exchange to signify who it is to another peer. The Identifier is
used by the receiver as an index into a table containing further
information about the peer that is required to continue processing
the message, for example a Security Association (SA) or keys.
Lebovitz, et al. Informational [Page 4]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
Identity Authentication
Once the identity is verified, there must be a cryptographic proof
of that identity, to ensure that the peer really is who it asserts
to be. Proof of identity can be arranged among peers in a few
ways, for example, symmetric and asymmetric pre-shared keys, or an
asymmetric key contained in a certificate. Certificates can be
used in ways that require no additional supporting systems
external to the routers themselves. An example of this is using
self-signed certificates and a flat file list of "approved
thumbprints". The different identity verification mechanisms vary
in ease of deployment, ease of ongoing management, startup effort,
security strength, and consequences from loss of secrets from one
part of the system to the rest of the system. For example, they
differ in resistance to a security breach, and the effort required
to recover in the event of such a breach. The point here is that
there are options, many of which are quite simple to employ and
deploy.
KDF (Key Derivation Function)
A KDF is a function in which an input key and other input data are
used to generate keying material that can be employed by
cryptographic algorithms. The key that is input to a KDF is
called a key derivation key. KDFs can be used to generate one or
more keys from (i) a random or pseudorandom seed value, or (ii)
the result of the Diffie-Hellman exchange, or (iii) a non-uniform
random source (e.g., from a non-deterministic random bit
generator), or (iv) a pre-shared key that may or may not be
memorable by a human.
KMP (Key Management Protocol)
KMP is a protocol that establishes a shared symmetric key between
a pair (or among a group) of users. It determines how secret keys
are made available to the users, and in some cases also determines
how the secret keys are generated. In some routing protocols, the
routing protocol derives the traffic keys from a master key. In
this case, KMP is responsible for the master-key generation and
for determining when the master key should be renewed. In other
cases, there are only traffic keys (and no master key); in such a
case, KMP is responsible for the traffic key generation and
renewal mechanism.
KMP Function
Any KMP used in the general KARP solution framework.
Peer Key
Peer keys are keys that are used among peers as a basis for
identifying one another. These keys may or may not be connection
specific, depending on how they were established, and what forms
Lebovitz, et al. Informational [Page 5]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
of identity and identity authentication mechanism are used in the
system. A peer key generally would be provided by a KMP and would
later be used to derive fresh traffic keys.
PSK (Pre-Shared Key)
A PSK is a key used to communicate with one or more peers in a
secure configuration. It is always distributed out of band prior
to a first connection.
Replayed Messages
Replayed messages are genuine messages that have been re-sent by
an attacker. Messages may be replayed within a session (i.e.,
intra-session) or replayed from a different session (i.e., inter-
session). For non-TCP-based protocols like OSPF [RFC2328] and
IS-IS [RFC1195], two routers are said to have a session up if they
are able to exchange protocol packets (i.e., the peers have an
adjacency). Messages replayed during an adjacency are intra-
session replays, while a message replayed between two peers who
re-establish an adjacency after a reboot or loss of connectivity
are inter-session replays.
Routing Protocol
This term refers to a Routing Protocol on which a KARP team is
working to improve the security of its packets on the wire.
SA (Security Association)
An SA is a relationship established between two or more entities
to enable them to protect the data they exchange. Examples of
attributes that may be associated with an SA include Identifier,
PSK, Traffic Key, cryptographic algorithms, and key lifetimes.
Threat Source
A threat source is a motivated, capable adversary.
Traffic Key
A Traffic Key is the key (or one of a set of keys) used for
protecting the routing protocol traffic. A traffic key should not
be a fixed value in a device configuration. A traffic key should
be known only to the participants in a connection, so that a
compromise of a stored key (possibly available to a terminated or
turned employee) does not result in disclosure of traffic keys.
If a server or other data store is stolen or compromised, the
attackers gain no access to current traffic keys. They may gain
access to key-derivation material, like a PSK, but not traffic
keys currently in use.
Additional terminology specific to threats are listed and defined
below in Section 3.
Lebovitz, et al. Informational [Page 6]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
1.2. Requirements Language
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [RFC2119].
When used in lower case, these words convey their typical use in
common language, and are not to be interpreted as described in RFC
2119.
2. KARP Effort Overview
2.1. KARP Scope
Three basic principles can be used to secure any piece of data as it
is transmitted over the wire: confidentiality, authenticity, and
integrity. The focus for the KARP working group will be message
authentication and message integrity only. At this time, this work
explicitly excludes confidentiality. Non-repudiation is also
excluded as a goal at this time. Since the objective of most routing
protocols is to broadly advertise the routing topology, routing
protocol packets are commonly sent in the clear; confidentiality is
not normally required for routing protocols. However, ensuring that
routing peers are authentically identified and that no rogue peers or
unauthenticated packets can compromise the stability of the routing
environment are critical and thus in scope. Confidentiality and non-
repudiation may be addressed in future work.
OSPF [RFC5709], IS-IS [RFC5310], LDP [RFC5036], and RIP [RFC2453]
[RFC4822] already incorporate mechanisms for cryptographically
authenticating and integrity checking the messages on the wire.
Products and code that incorporate these mechanisms have been
produced and have been optimized for these existing security
mechanisms. Rather than turn away from these mechanisms, this
document aims to enhance them, updating them to modern and more
secure levels.
Therefore, the scope of KARP's roadmap of work includes:
o Making use of existing routing protocol transport security
mechanisms, where they have been specified, and enhancing or
updating them as necessary for modern cryptographic best
practices. [RFC6518], Section 4.1 labels this KARP's Work Phase 1.
o Developing a framework for using automatic key management in order
to ease deployment, lower cost of operation, and allow for rapid
responses to security breaches. [RFC6518], Section 4.1 labels
this KARP's Work Phase 2.
Lebovitz, et al. Informational [Page 7]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
o Specifying an automated key management protocol that may be
combined with Routing Protocol mechanisms. [RFC6518], Section 4.1
labels this KARP's Work Phase 2.
Neither this document nor [RFC6518] contains protocol specifications.
Instead, they define the areas in which protocol specification work
is needed, and they set a direction, a set of requirements, and
priorities for addressing that specification work.
There are a set of threats to routing protocols that are considered
in scope for KARP, and a set considered out of scope. These are
described in detail in Section 3.
2.2. Incremental Approach
This document serves as an agreement between the Routing Area and the
Security Area about the priorities and work plan for incrementally
delivering the work described in the KARP roadmap above. The
principle of "crawl, walk, run" will be employed. Thus routing
protocol authentication mechanisms may not go immediately from their
current state to a state reflecting the best possible, most modern
security practices. This point is important as there will be times
when the best security possible will give way to security that is
vastly improved over current security but that is admittedly not the
best security possible, in order that incremental progress toward a
more secure Internet may be achieved. As such, this document will
call out places where agreement has been reached on such trade-offs.
Incremental steps will need to be taken for a few very practical
reasons. First, there are a considerable number of deployed routing
devices in operating networks that will not be able to run the most
modern cryptographic mechanisms without significant and unacceptable
performance penalties. The roadmap for any routing protocol MUST
allow for incremental improvements on existing operational devices.
Second, current routing protocol performance on deployed devices has
been achieved over the last 20 years through extensive tuning of
software and hardware elements, and is a constant focus for
improvement by vendors and operators alike. The introduction of new
security mechanisms affects this performance balance. The
performance impact of any incremental security improvement will need
to be weighed by the community and introduced in such a way that
allows the vendor and operator community a path to adoption that
upholds reasonable performance metrics. Therefore, certain
specification elements may be introduced carrying the "SHOULD"
guidance, with the intention that the same mechanism will carry a
"MUST" in a future release of the specification. This approach gives
the vendors and implementors the guidance they need to tune their
software and hardware appropriately over time. Last, some security
Lebovitz, et al. Informational [Page 8]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
mechanisms require the build-out of other operational support
systems, which will take time.
An example where these three steps were at play in an incremental
improvement roadmap was the improvement of BGP's [RFC4271] security
via the TCP Authentication Option (TCP-AO) [RFC5925] effort. It
would have been ideal, and would have reflected best common security
practice, to have a fully specified key management protocol for
negotiating the TCP-AO keying material, e.g., using certificates for
peer authentication. However, in the spirit of incremental
deployment, the IETF first addressed issues like cryptographic
algorithm agility, replay attacks, and the resetting of TCP sessions
in the base TCP-AO protocol, and then later began work to layer key
management on top of these.
2.3. Goals
The goals and general guidance for the KARP work follow:
1. Provide authentication and integrity protection for messages on
the wire for existing routing protocols.
2. Define a path to incrementally improve security of the routing
infrastructure as explained in Section 2.2.
3. Ensure that the improved security solutions are deployable on
current routing infrastructure. This requires consideration of
the current state of processing power available on routers in the
network today.
4. Operational deployability - A solution's acceptability also will
be measured by how deployable the solution is by operator teams,
with consideration for their deployment processes and
infrastructures. Specifically, KARP design teams will try to
make these solutions fit as well as possible into current
operational practices and router deployment methodologies. Doing
so will depend heavily on operator input during KARP design
efforts. Hopefully, operator input will lead to a more
deployable solution, which will, in turn, lead to more production
deployments. Deployment of incrementally more secure routing
infrastructure in the Internet is the final measure of success.
We would like to see an increase in the number of respondents to
surveys such as [ISR2008] to report deployment of the updated
authentication and integrity mechanisms in their networks, as
well as see a sharp rise in usage of these mechanisms across a
greater percentage of their network's routers.
Lebovitz, et al. Informational [Page 9]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
Interviews with operators show several points about routing
security. First, according to [ISR2008], over 70% of operators
have deployed transport connection protection via TCP MD5
[RFC3562] on their External Border Gateway Protocol (eBGP)
sessions. Over 55% also deploy TCP MD5 on their Internal Border
Gateway Protocol (iBGP) connections, and 50% make use of TCP MD5
offered on some other internal gateway protocol (IGP). The same
survey states that "a considerable increase was observed over
previous editions of the survey for use of TCP MD5 with external
peers (eBGP), internal peers (iBGP) and MD5 extensions for IGPs."
Though the data is not captured in the report, the authors
believe anecdotally that of those who have deployed TCP MD5
somewhere in their network, only about 25-30% of the routers in
their network are deployed with the authentication enabled. None
report using IPsec [RFC4301] to protect the routing protocol,
which was a decline from the few that reported doing so in the
previous year's report. Anecdotal evidence from operators using
MD5 shows that almost all report using one manually distributed
key throughout the entire network. These same operators report
that the single key has not been changed since it was originally
installed, sometimes five or more years ago. When asked why,
particularly for the case of protecting BGP sessions using TCP
MD5, the following reasons were often given:
A. Changing the keys triggers a TCP reset, and thus the links/
adjacencies bounce, undermining Service Level Agreements
(SLAs).
B. For external peers, it is difficult to coordinate with the
other organization, and in practice the coordination is very
cumbersome and tedious to execute. Once the operator finds
the correct contact at the other organization (not always so
easy), the coordination function is serialized and performed
on a per-peer or per-AS basis.
C. Keys must be changed at precisely the same time, or at least
within 60 seconds (as supported by two major vendors) in order
to limit the duration of a connectivity outage. This is
incredibly difficult to do, operationally, especially between
different organizations.
D. Key change is perceived as a relatively low priority compared
to other operational issues.
E. Staff levels are insufficient to implement the changes on a
device-by-device basis.
Lebovitz, et al. Informational [Page 10]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
F. There are three use cases for operational peering at play:
peers and interconnection with other operators, iBGP and other
routing sessions within a single operator, and operator-to-
customer devices. All three have very different properties,
and all are reported as cumbersome to manage securely. One
operator reported that the same key is used for all customer
premise equipment (CPE). The same operator reported that if
the customer mandated it, a unique key could be created,
although the last time this occurred, it created such an
operational headache that the administrators now usually tell
customers that the option doesn't even exist, to avoid the
difficulties. These customer-unique keys are never changed,
unless the customer demands so. The main threat here is that
a terminated employee from such an operator who had access to
the one (or several) keys used for authentication in these
environments could wage an attack. Alternatively, the
operator could offer the keys to others who would wage the
attack. In either case, the attacker could then bring down
many of the adjacencies, thus destabilizing the routing
system.
5. Whatever mechanisms KARP specifies need to be easier to deploy
than the current methods and should provide obvious operational
efficiency gains along with significantly better security. This
combination of value may be enough to drive much broader
adoption.
6. Address the threats enumerated below in "Threats" (Section 3) for
each routing protocol. Not all threats may be able to be
addressed in the first specification update for any one protocol.
Roadmaps will be defined so that both the Security Area and the
Routing Area agree on how the threats will be addressed
completely over time.
7. Create a reusable architecture, framework, and guidelines for
various IETF working groups that will address these security
improvements for various Routing Protocols. The crux of the KARP
work is to reuse the architecture, framework, and guidelines as
much as possible across relevant Routing Protocols. For example,
designers should aim to reuse the key management protocol that
will be defined for BGP, which will establish keys for TCP-AO,
for as many other routing protocols with similar characteristics
and properties as possible.
8. Bridge any gaps between the IETF Routing and Security Areas by
recording agreements on work items, roadmaps, and guidance from
the cognizant Area Directors and the Internet Architecture Board
(IAB).
Lebovitz, et al. Informational [Page 11]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
2.4. Non-Goals
The following goals are considered out of scope for this effort:
o Confidentiality and non-repudiation of the packets on the wire.
Once the goals of this roadmap are realized, work on
confidentiality may be considered.
o Non-repudiation of the packets on the wire.
o Message content validity (routing database validity). This work
is being addressed in other IETF efforts. For example, BGP
message content validity is being addressed in the SIDR working
group.
2.5. Audience
The audience for this document includes:
o Routing Area working group chairs and participants - These people
are charged with updating Routing Protocol specifications. Any
and all cryptographic authentication work on these specifications
will occur in Routing Area working groups, in close partnership
with the Security Area. Co-advisors from the Security Area may
often be named for these partnership efforts.
o Security Area reviewers of Routing Area documents - These people
are tasked by the Security Area Directors to perform reviews on
routing protocol specifications as they pass through working group
last call or IESG review. Their particular attention to the use
of cryptographic authentication and newly specified security
mechanisms for the routing protocols is appreciated. They also
help to ensure that incremental security improvements are being
made, in line with this roadmap.
o Security Area engineers - These people partner with Routing Area
authors/designers on the security mechanisms in routing protocol
specifications. Some of these Security Area engineers will be
assigned by the Security Area Directors, while others will be
interested parties in the relevant working groups.
o Operators - The operators are a key audience for this work, as the
work is considered to have succeeded only if operators deploy the
technology. It is anticipated that deployment will take place
only if operators perceive that the improved security offered by
the Routing Protocol updates warrants the complexity and cost of
deployment and operation. Conversely, the work will be considered
a failure if operators do not deploy it, either due to a lack of
Lebovitz, et al. Informational [Page 12]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
perceived value or due to perceived operational complexity. As a
result, the GROW and OPSEC working groups should be kept squarely
in the loop as well.
3. Threats
This document uses the definition of "threat" from RFC 4949
[RFC4949]: "[a] potential for violation of security, which exists
when there is an entity, circumstance, capability, action, or event
that could cause harm."
This section defines the threats that are in scope for the KARP
effort. It also lists those threats that are explicitly out of scope
for the KARP effort. Threats are discussed assuming that no
protection (i.e., message authentication and message integrity) has
been applied to routing protocol messages.
This document leverages the model described in "Generic Threats to
Routing Protocols" [RFC4593]. Specifically, the threats listed below
were derived by reviewing [RFC4593], analyzing how the threats
applied to the KARP problem space, and listing the threats that are
applicable to the work for the KARP design team. This document
categorizes [RFC4593] threats into those in scope and those out of
scope for KARP. Each in-scope threat is discussed below, and its
applicability to the KARP problem space is described. As such, the
following text intentionally is not a comprehensive threat analysis.
Rather, it describes the applicability of the existing threat
analysis in [RFC4593] to KARP.
Note: terms from [RFC4593] appear capitalized below -- e.g.
OUTSIDERS -- so as to make explicit the term's origin, and to enable
rapid cross referencing to the source RFC.
For convenience, a terse definition of most [RFC4593] terms is
offered here. Those interested in a more thorough description of
routing protocol threat sources, motivations, consequences, and
actions will want to read [RFC4593] before continuing here.
3.1. Threat Sources
3.1.1. OUTSIDERS
One of the threats that will be addressed in this roadmap is the
situation in which the source is an OUTSIDER. An OUTSIDER attacker
may reside anywhere in the Internet, may have the ability to send IP
traffic to the router, may be able to observe the router's replies,
and may even control the path for a legitimate peer's traffic.
OUTSIDERS are not legitimate participants in the routing protocol.
Lebovitz, et al. Informational [Page 13]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
The use of message authentication and integrity protection
specifically aims to identify packets originating from OUTSIDERS.
KARP design teams will consider two specific use cases of OUTSIDERS:
those on path, and those off path.
o On Path - These attackers have control of a network resource or a
tap that sits along the path between two routing peers. A "Man in
the Middle" (MitM) is an on-path attacker. From this vantage
point, the attacker can conduct either active or passive attacks.
An active attack occurs when the attacker places packets on the
network as part of the attack. One active MitM attack relevant to
KARP, an active wiretapping attack, occurs when the attacker
tampers with packets moving between two legitimate router peers in
such a way that both peers think they are talking to each other
directly, when in fact they are actually talking to the attacker.
Protocols conforming to this roadmap will use cryptographic
mechanisms to detect MitM attacks and reject packets from such
attacks (i.e., discard them as being not authentic). Passive on-
path attacks occur when the attacker silently gathers data and
analyzes it to gain advantage. Passive activity by an on-path
attacker may lead to an active attack.
o Off Path - These attackers sit on some network outside of that
over which the packets between two routing peers run. The source
may be one or several hops away. Off-path attackers can launch
active attacks, such as SPOOFING or denial-of-service (DoS)
attacks, to name a few.
3.1.2. Unauthorized Key Holder
This threat source exists when an unauthorized entity somehow manages
to gain access to keying material. Using this material, the attacker
could send packets that pass the authenticity checks based on Message
Authentication Codes (MACs). The resulting traffic might appear to
come from router A and be destined for router B, and thus the
attacker could impersonate an authorized peer. The attacker could
then adversely affect network behavior by sending bogus messages that
appear to be authentic. The attack source possessing the
unauthorized keys could be on path, off path, or both.
The obvious mitigation for an unauthorized key holder is to change
the keys currently in use by the legitimate routing peers. This
mitigation can be either reactive or proactive. Reactive mitigation
occurs when keys are changed only after one has discovered that the
previous keys have fallen into the possession of unauthorized users.
The reactive mitigation case is highlighted here in order to explain
a common operational situation where new keying material will need to
Lebovitz, et al. Informational [Page 14]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
be put in place with little or no advanced warning. In such a case,
new keys must be able to be installed and put into use very quickly,
and with little operational expense. Proactive mitigation occurs
when an operator assumes that unauthorized possession will occur from
time to time without being discovered, and the operator moves to new
keying material in order to cut short an attacker's window of
opportunity to use the stolen keys effectively.
KARP design teams can address this type of attack by creating
specifications that make it practical for the operator to quickly
change keys without disruption to the routing system and with minimal
operational overhead. Operators can further mitigate threats from
unauthorized key holders by regularly changing keys.
3.1.2.1. Terminated Employee
A terminated employee is an important example of an unauthorized key
holder. Staff attrition is a reality in routing operations and is
therefore a potential threat source. The threat source risk arises
when a network operator who had been granted access to keys ceases to
be an employee. If new keys are deployed immediately, the situation
of a terminated employee can become an "unauthorized key holder,
proactive" case, as described above, rather than an "unauthorized key
holder, reactive mitigation" case. It behooves the operator to
change the keys, to enforce the revocation of authorization of the
old keys, in order to minimize the threat source's window of
opportunity.
A terminated employee is a valid unauthorized key holder threat
source for KARP, and designs should address the associated threats.
For example, new keys must be able to be installed and made
operational in the routing protocols very quickly, with zero impact
to the routing system, and with little operational expense. The
threat actions associated with a terminated employee also motivate
the need to change the keys quickly, also with little operational
expense.
3.1.3. BYZANTINE
According to [RFC4593], Section 3.1.1.2, BYZANTINE "attackers are
faulty, misconfigured, or subverted routers; i.e., legitimate
participants in the routing protocol", whose messages cause routing
to malfunction.
[RFC4593] goes on to say that "[s]ome adversaries can subvert
routers, or the management workstations used to control these
routers. These Byzantine failures represent the most serious form of
Lebovitz, et al. Informational [Page 15]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
attack capability in that they result in emission of bogus traffic by
legitimate routers."
[RFC4593] explains that "[d]eliberate attacks are mimicked by
failures that are random and unintentional. In particular, a
Byzantine failure in a router may occur because the router is faulty
in hardware or software or is misconfigured", and thus routing
malfunctions unintentionally. Although not malicious, such
occurrences still disrupt network operation.
Whether faulty, misconfigured, or subverted, Byzantine routers have
an empowered position from which to provide believable yet bogus
routing messages that are damaging to the network.
3.2. Threat Actions In Scope
The following THREAT ACTIONS are in scope for KARP:
o SPOOFING - when an unauthorized device assumes the identity of an
authorized one. Spoofing is special in that it can be used to
carry out other threat actions that cause other threat
consequences. SPOOFING can be used, for example, to inject
malicious routing information that causes the disruption of
network services. SPOOFING can also be used to cause a neighbor
relationship to form that subsequently denies the formation of the
relationship with a legitimate router.
o DoS attacks
A. At the transport layer - This occurs when an attacker sends
packets aimed at halting or preventing the underlying protocol
over which the routing protocol runs. The attacker could use
SPOOFING, FALSIFICATION, or INTERFERENCE (see below) to
produce the DoS attack. For example, BGP running over
Transport Layer Security (TLS) will still not solve the
problem of an attacker being able to send a spoofed TCP FIN or
TCP RST and causing the BGP session to go down. Since these
attacks depend on spoofing, operators are encouraged to deploy
proper authentication mechanisms to prevent them.
Specification work should ensure that Routing Protocols can
operate over transport subsystems in a fashion that is
resilient to such DoS attacks.
B. Using the authentication mechanism - This includes an attacker
causing INTERFERENCE, which inhibits exchanges of legitimate
routers. The attack is often perpetrated by sending packets
that confuse or overwhelm a security mechanism itself. An
example is initiating an overwhelming load of spoofed routing
Lebovitz, et al. Informational [Page 16]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
protocol packets that contain a MAC (i.e., INSERTING
MESSAGES), so that the receiver spends substantial CPU
resources on the processing cycles to check the MAC, only to
discard the spoofed packet. Other types of INTERFERENCE
include REPLAYING OUT-DATED PACKETS, CORRUPTING MESSAGES, and
BREAKING SYNCHRONIZATION.
o FALSIFICATION - An action whereby an attacker sends false routing
information. This document targets only FALSIFICATION from
OUTSIDERS that may occur from tampering with packets in flight or
sending entirely false messages. FALSIFICATION from BYZANTINES
(see Section 3.3) are not addressed by the KARP effort.
o Brute-Force Attacks Against Password/Keys - This includes either
online or offline attacks in which attempts are made repeatedly
using different keys/passwords until a match is found. While it
is impossible to make brute-force attacks on keys completely
unsuccessful, proper design can make it much harder for such
attacks to succeed. For example, current guidance for the
security strength of an algorithm with a particular key length
should be deemed acceptable for a period of 10 years. (Section 10
of [SP.800-131A] is one source for guidance.) Using per-session
keys is another widely used method for reducing the number of
brute-force attacks, as this would make it difficult to guess the
keys.
3.3. Threat Actions Out of Scope
BYZANTINE sources -- be they faulty, misconfigured, or subverted --
are out of scope for this roadmap. KARP works to cryptographically
ensure that received routing messages originated from authorized
peers and that the message was not altered in transit. Formation of
a bogus message by a valid and authorized peer falls outside the KARP
scope. Any of the attacks described in Section 3.2 that may be
levied by a BYZANTINE source are therefore also out of scope, e.g.
FALSIFICATION from BYZANTINE sources or unauthorized message content
by a legitimate authorized peer.
In addition, these other attack actions are out of scope for this
work:
o SNIFFING (passive wiretapping) - Passive observation of route
message contents in flight. Data confidentiality, as achieved by
data encryption, is the common mechanism for preventing SNIFFING.
While useful, especially to prevent the gathering of data needed
to perform an off-path packet injection attack, data encryption is
out of scope for KARP.
Lebovitz, et al. Informational [Page 17]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
o INTERFERENCE due to:
A. NOT FORWARDING PACKETS - Cannot be prevented with
cryptographic authentication. Note: If sequence numbers with
sliding windows are used in the solution (as is done, for
example, in Bidirectional Forwarding Detection (BFD)
[RFC5880]), a receiver can at least detect the occurrence of
this attack.
B. DELAYING MESSAGES - Cannot be prevented with cryptographic
authentication. Note: Timestamps can be used to detect
delays.
C. DENIAL OF RECEIPT (non-repudiation) - Cannot be prevented with
cryptographic authentication.
D. UNAUTHORIZED MESSAGE CONTENT - Covered by the work of the
IETF's SIDR working group
(http://www.ietf.org/html.charters/sidr-charter.html).
E. DoS attacks not involving the routing protocol. For example,
a flood of traffic that fills the link ahead of the router, so
that the router is rendered unusable and unreachable by valid
packets is NOT an attack that KARP will address. Many such
examples could be contrived.
4. Requirements for KARP Work Phase 1: Update to a Routing Protocol's
Existing Transport Security
Section 4.1 of the KARP Design Guide [RFC6518] describes two distinct
work phases for the KARP effort. This section addresses requirements
for the first work phase only, Work Phase 1, the update to a routing
protocol's existing transport security. Work Phase 2, the framework
and usage of a KMP, will be addressed in a future document(s).
The following list of requirements SHOULD be addressed by a KARP Work
Phase 1 security update to any Routing Protocol (according to section
4.1 of the KARP Design Guide [RFC6518]document). IT IS RECOMMENDED
that any Work Phase 1 security update to a Routing Protocol contain a
section of the specification document that describes how each of the
following requirements are met. It is further RECOMMENDED that
justification be presented for any requirements that are NOT
addressed.
1. Clear definitions of which elements of the transmitted data
(frame, packet, segment, etc.) are protected by an
authentication/integrity mechanism.
Lebovitz, et al. Informational [Page 18]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
2. Strong cryptographic algorithms, as defined and accepted by the
IETF security community, MUST be specified. The use of non-
standard or unpublished algorithms MUST be avoided.
3. Algorithm agility for the cryptographic algorithms used in the
authentication MUST be specified, and protocol specifications
MUST be clear regarding how new algorithms are specified and
used within the protocol. This requirement exists because
research identifying weaknesses in cryptographic algorithms can
cause the security community to reduce confidence in some
algorithms. Breaking a cipher isn't a matter of if, but when it
will occur. Having the ability to specify alternate algorithms
(algorithm agility) within the protocol specification to support
such an event is essential. Additionally, more than one
algorithm MUST be specified. Mandating support for two
algorithms (i.e., one mandatory to implement algorithm and one
or more backup algorithms to guide transition) provides both
redundancy, and a mechanism for enacting that redundancy.
4. Secure use of PSKs, offering both operational convenience and a
baseline level of security, MUST be specified.
5. Routing Protocols (or the transport or network mechanism
protecting routing protocols) SHOULD be able to detect and
reject replayed intra-session and inter-session messages.
Packets captured from one session MUST NOT be able to be resent
and accepted during a later session (i.e., inter-session
replay). Additionally, replay mechanisms MUST work correctly
even in the presence of routing protocol packet prioritization
by the router.
There is a specific case of replay attack combined with spoofing
that must be addressed. Several routing protocols (e.g., OSPF
[RFC2328], IS-IS [RFC1195], BFD [RFC5880], RIP [RFC2453], etc.),
require all speakers to share the same authentication and
message association key on a broadcast segment. It is important
that an integrity check associated with a message fail if an
attacker has replayed the message with a different origin.
6. A change of security parameters MUST force a change of session
traffic keys. The specific security parameters for the various
routing protocols will differ and will be defined by each
protocol design team. Some examples may include master key, key
lifetime, and cryptographic algorithm. If one of these
configured parameters changes, then a new session traffic key
MUST immediately be established using the updated parameters.
The routing protocol security mechanisms MUST support this
behavior.
Lebovitz, et al. Informational [Page 19]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
7. Security mechanisms MUST specify a means to affect intra-session
rekeying without disrupting a routing session. This should be
accomplished without data loss, if possible. Keys may need to
be changed periodically based on policy or when an administrator
who had access to the keys leaves an organization. A rekeying
mechanism enables the operators to execute the change without
productivity loss.
8. Rekeying SHOULD be supported in such a way that it can occur
during a session without the peer needing to use multiple keys
to validate a given packet. The rare exception will occur if a
routing protocol's design team can find no other way to rekey
and still adhere to the other requirements in this section. The
specification SHOULD include a key identifier, which allows
receivers to choose the correct key (or determine that they are
not in possession of the correct key).
9. New mechanisms MUST resist DoS attacks described as in scope in
Section 3.2. Routers protect the control plane by implementing
mechanisms to reject completely or rate-limit traffic not
required at the control-plane level (i.e., unwanted traffic).
Typically, line-rate packet-filtering capabilities look at
information in the IP and transport (TCP or UDP) headers, but do
not include higher-layer information. Therefore, the new
mechanisms should neither hide nor encrypt the information
carried in the IP and transport layers in control-plane packets.
10. Mandatory cryptographic algorithms and mechanisms MUST be
specified for each routing protocol security mechanism.
Further, the protocol specification MUST define default security
mechanism settings for all implementations to use when no
explicit configuration is provided. To understand the need for
this requirement, consider the case where a routing protocol
mandates three different cryptographic algorithms for a MAC
operation. If company A implements algorithm 1 as the default
for this protocol, while company B implements algorithm 2 as the
default, then two operators who enable the security mechanism
with no explicit configuration other than a PSK will experience
a connection failure. It is not enough that each implementation
implement the three mandatory algorithms; one default must
further be specified in order to gain maximum out-of-the-box
interoperability.
11. For backward-compatibility reasons, manual keying MUST be
supported.
12. The specification MUST consider and allow for future use of a
KMP.
Lebovitz, et al. Informational [Page 20]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
13. The authentication mechanism in a Routing Protocol MUST be
decoupled from the key management system used. The
authentication protocol MUST include a specification for
agreeing on keying material. This will accommodate both manual
keying and the use of KMPs.
14. Convergence times of the Routing Protocols SHOULD NOT be
materially affected. Changes in the convergence time will be
immediately and independently verifiable by convergence
performance test beds already in use (e.g. those maintained by
router vendors, service providers, and researchers). An
increase in convergence time in excess of 5% is likely to be
considered to have materially affected convergence by network
operators. A number of other factors can also change
convergence over time (e.g., speed of processors used on
individual routing peers, processing power increases due to
Moore's law, and implementation specifics), and implementors
will need to take into account the effect of an authentication
mechanism on Routing Protocols. Protocol designers should
consider the impact on convergence times as a function of both
the total number of protocol packets that must be exchanged and
the required computational processing of individual messages in
the specification, understanding that the operator community's
threshold for an increase in convergence times is very low, as
stated above.
15. The changes to or addition of security mechanisms SHOULD NOT
cause a refresh of route advertisements or cause additional
route advertisements to be generated.
16. Router implementations provide prioritized treatment for certain
protocol packets. For example, OSPF Hello and Acknowledgement
packets are prioritized for processing above other OSPF packets.
The security mechanism SHOULD NOT interfere with the ability to
observe and enforce such prioritization. Any effect on such
priority mechanisms MUST be explicitly documented and justified.
Replay protection mechanisms provided by the routing protocols
MUST work even if certain protocol packets are offered
prioritized treatment.
17. The Routing Protocol MUST send minimal information regarding the
authentication mechanisms and associated parameters in its
protocol packets. This keeps the Routing Protocols as clean and
focused as possible, and loads security negotiations into the
KMP as much as possible. This also avoids exposing any security
negotiation information unnecessarily to possible attackers on
the path.
Lebovitz, et al. Informational [Page 21]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
18. Routing Protocols that rely on the IP header (or information
separate from routing protocol payload) to identify the neighbor
that originated the packet MUST either protect the IP header or
provide some other means to authenticate the neighbor.
[RFC6039] describes some attacks that motivate this requirement.
19. Every new KARP-developed security mechanisms MUST support
incremental deployment. It will not be feasible to deploy a new
Routing Protocol authentication mechanism throughout a network
instantaneously. Indeed, it may not actually be feasible to
deploy such a mechanism to all routers in a large autonomous
system (AS) in a bounded timeframe. Proposed solutions MUST
support an incremental deployment method that benefits those who
participate. Because of this, there are several requirements
that any proposed KARP mechanism should consider.
A. The Routing Protocol security mechanism MUST enable each
router to configure use of the security mechanism on a per-
peer basis where the communication is peer to peer
(unicast).
B. Every new KARP-developed security mechanism MUST provide
backward compatibility with respect to message formatting,
transmission, and processing of routing information carried
through secure and non-secure security environments.
Message formatting in a fully secured environment MAY be
handled in a non-backward-compatible fashion, though care
must be taken to ensure that routing protocol packets can
traverse intermediate routers that don't support the new
format.
C. In an environment where both secured and non-secured routers
are interoperating, a mechanism MUST exist for secured
systems to identify whether a peer intended the messages to
be secured.
D. In an environment where secured service is in the process of
being deployed, a mechanism MUST exist to support a
transition free of service interruption (caused by the
deployment per se).
20. The introduction of mechanisms to improve routing security may
increase the processing performed by a router. Since most of
the currently deployed routers do not have hardware to
accelerate cryptographic operations, these operations could
impose a significant processing burden under some circumstances.
Thus, proposed solutions SHOULD be evaluated carefully with
regard to the processing burden they may impose, since
Lebovitz, et al. Informational [Page 22]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
deployment may be impeded if network operators perceive that a
solution will impose a processing burden that either incurs
substantial capital expense or threatens to degrade router
performance.
21. New authentication and security mechanisms should not rely on
systems external to the routing system (the equipment that is
performing forwarding) in order for the routing system to be
secure. In order to ensure the rapid initialization and/or
return to service of failed nodes, it is important to reduce
reliance on these external systems to the greatest extent
possible. Proposed solutions SHOULD NOT require connections to
external systems, beyond those directly involved in peering
relationships, in order to return to full service. It is,
however, acceptable for the proposed solutions to require post-
initialization synchronization with external systems in order to
fully synchronize security associations.
If authentication and security mechanisms rely on systems
external to the routing system, then there MUST be one or more
options available to avoid circular dependencies. It is not
acceptable to have a routing protocol (e.g., unicast routing)
depend upon correct operation of a security protocol that, in
turn, depends upon correct operation of the same instance of
that routing protocol (i.e., the unicast routing). However, it
is acceptable to have operation of a routing protocol (e.g.,
multicast routing) depend upon operation of a security protocol,
which depends upon an independent routing protocol (e.g.,
unicast routing). Similarly, it would be okay to have the
operation of a routing protocol depend upon a security protocol,
which in turn uses an out-of-band network to exchange
information with remote systems.
5. Security Considerations
This document is mostly about security considerations for the KARP
efforts, both threats and the requirements for addressing those
threats. More detailed security considerations are provided in the
Security Considerations section of the KARP Design Guide
[RFC6518]document.
The use of a group key between a set of Routing Protocol peers has
special security considerations. Possession of the group key itself
is used for identity validation; no other identity check is used.
Under these conditions, an attack exists when one peer masquerades as
a neighbor by using the neighbor's source IP address. This type of
attack has been well documented in the group-keying problem space,
and it is non-trivial to solve. Solutions exist within the group-
Lebovitz, et al. Informational [Page 23]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
keying realm, but they come with significant increases in complexity
and computational intensity.
6. Acknowledgements
The majority of the text for initial draft of this document was taken
from "Roadmap for Cryptographic Authentication of Routing Protocol
Packets on the Wire", authored by Gregory M. Lebovitz.
Brian Weis provided significant assistance in handling the many
comments that came back during IESG review, including making textual
edits directly to the XML. For his extensive efforts he was added as
an author.
We would like to thank the following people for their thorough
reviews and comments: Brian Weis, Yoshifumi Nishida, Stephen Kent,
Vishwas Manral, Barry Leiba, Sean Turner, and Uma Chunduri.
Author Gregory M. Lebovitz was employed at Juniper Networks, Inc. for
much of the time he worked on this document, though not at the time
of its publishing. Thus, Juniper sponsored much of this effort.
7. References
7.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC4593] Barbir, A., Murphy, S., and Y. Yang, "Generic Threats
to Routing Protocols", RFC 4593, October 2006.
[RFC4948] Andersson, L., Davies, E., and L. Zhang, "Report from
the IAB workshop on Unwanted Traffic March 9-10,
2006", RFC 4948, August 2007.
7.2. Informative References
[ISR2008] McPherson, D. and C. Labovitz, "Worldwide
Infrastructure Security Report", October 2008,
<http://pages.arbornetworks.com/rs/arbor/images/
ISR2008_EN.pdf>.
[RFC1195] Callon, R., "Use of OSI IS-IS for routing in TCP/IP
and dual environments", RFC 1195, December 1990.
[RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328,
April 1998.
Lebovitz, et al. Informational [Page 24]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
[RFC2453] Malkin, G., "RIP Version 2", STD 56, RFC 2453,
November 1998.
[RFC3562] Leech, M., "Key Management Considerations for the TCP
MD5 Signature Option", RFC 3562, July 2003.
[RFC4271] Rekhter, Y., Ed., Li, T., Ed., and S. Hares, Ed., "A
Border Gateway Protocol 4 (BGP-4)", RFC 4271,
January 2006.
[RFC4301] Kent, S. and K. Seo, "Security Architecture for the
Internet Protocol", RFC 4301, December 2005.
[RFC4822] Atkinson, R. and M. Fanto, "RIPv2 Cryptographic
Authentication", RFC 4822, February 2007.
[RFC4949] Shirey, R., "Internet Security Glossary, Version 2",
FYI 36, RFC 4949, August 2007.
[RFC5036] Andersson, L., Ed., Minei, I., Ed., and B. Thomas,
Ed., "LDP Specification", RFC 5036, October 2007.
[RFC5310] Bhatia, M., Manral, V., Li, T., Atkinson, R., White,
R., and M. Fanto, "IS-IS Generic Cryptographic
Authentication", RFC 5310, February 2009.
[RFC5709] Bhatia, M., Manral, V., Fanto, M., White, R., Barnes,
M., Li, T., and R. Atkinson, "OSPFv2 HMAC-SHA
Cryptographic Authentication", RFC 5709, October 2009.
[RFC5880] Katz, D. and D. Ward, "Bidirectional Forwarding
Detection (BFD)", RFC 5880, June 2010.
[RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP
Authentication Option", RFC 5925, June 2010.
[RFC6039] Manral, V., Bhatia, M., Jaeggli, J., and R. White,
"Issues with Existing Cryptographic Protection Methods
for Routing Protocols", RFC 6039, October 2010.
[RFC6518] Lebovitz, G. and M. Bhatia, "Keying and Authentication
for Routing Protocols (KARP) Design Guidelines",
RFC 6518, February 2012.
Lebovitz, et al. Informational [Page 25]
RFC 6862 KARP Overview, Threats, and Requirements March 2013
[SP.800-131A] Barker, E. and A. Roginsky, "Transitions:
Recommendation for Transitioning the Use of
Cryptographic Algorithms and Key Lengths", United
States of America, National Institute of Science and
Technology, NIST Special Publication 800-131A,
January 2011.
Authors' Addresses
Gregory Lebovitz
Aptos, California 95003
United States
EMail: gregory.ietf@gmail.com
Manav Bhatia
Alcatel-Lucent
Bangalore,
India
EMail: manav.bhatia@alcatel-lucent.com
Brian Weis
Cisco Systems
170 W. Tasman Drive
San Jose, California 95134-1706
United States
EMail: bew@cisco.com
URI: http://www.cisco.com
Lebovitz, et al. Informational [Page 26]