Internet Engineering Task Force (IETF) D. Dugal
Request for Comments: 6192 Juniper Networks
Category: Informational C. Pignataro
ISSN: 2070-1721 R. Dunn
Cisco Systems
March 2011
Protecting the Router Control Plane
Abstract
This memo provides a method for protecting a router's control plane
from undesired or malicious traffic. In this approach, all
legitimate router control plane traffic is identified. Once
legitimate traffic has been identified, a filter is deployed in the
router's forwarding plane. That filter prevents traffic not
specifically identified as legitimate from reaching the router's
control plane, or rate-limits such traffic to an acceptable level.
Note that the filters described in this memo are applied only to
traffic that is destined for the router, and not to all traffic that
is passing through the router.
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/rfc6192.
Copyright Notice
Copyright (c) 2011 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
Dugal, et al. Informational [Page 1]
RFC 6192 Protect Router Control Plane March 2011
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 ....................................................2
2. Applicability Statement .........................................4
3. Method ..........................................................4
3.1. Legitimate Traffic .........................................5
3.2. Filter Design ..............................................6
3.3. Design Trade-Offs ..........................................7
3.4. Additional Protection Considerations ......................10
4. Security Considerations ........................................10
5. Acknowledgements ...............................................11
6. Informative References .........................................12
Appendix A. Configuration Examples ................................13
A.1. Cisco Configuration .......................................13
A.2. Juniper Configuration .....................................17
1. Introduction
Modern router architecture design maintains a strict separation of
forwarding and router control plane hardware and software. The
router control plane supports routing and management functions. It
is generally described as the router architecture hardware and
software components for handling packets destined to the device
itself as well as building and sending packets originated locally on
the device. The forwarding plane is typically described as the
router architecture hardware and software components responsible for
receiving a packet on an incoming interface, performing a lookup to
identify the packet's IP next hop and determine the best outgoing
interface towards the destination, and forwarding the packet out
through the appropriate outgoing interface.
Visually, this architecture can be represented as the router's
control plane hardware sitting on top of, and interfacing with, the
forwarding plane hardware with interfaces connecting to other network
devices. See Figure 1.
Dugal, et al. Informational [Page 2]
RFC 6192 Protect Router Control Plane March 2011
+----------------+
| Router Control |
| Plane |
+------+ +-------+
| |
Router Control
Plane Protection
| |
+------+ +-------+
| Forwarding |
Interface X ==[ Plane ]== Interface Y
+----------------+
Figure 1: Router Control Plane Protection
Typically, forwarding plane functionality is realized in high-
performance Application Specific Integrated Circuits (ASICs) that are
capable of handling very high packet rates. By contrast, the router
control plane is generally realized in software on general-purpose
processors. While software instructions run on both planes, the
router control plane hardware is usually not optimized for high-speed
packet handling. Given their differences in packet-handling
capabilities, the router's control plane hardware is more susceptible
to being overwhelmed by a Denial-of-Service (DoS) attack than the
forwarding plane's ASICs. It is imperative that the router control
plane remain stable regardless of traffic load to and from the device
because the router control plane is what drives the programming of
the forwarding plane.
The router control plane also processes traffic destined to the
router, and because of the wider range of functionality is more
susceptible to security vulnerabilities and a more likely target for
a DoS attack than the forwarding plane.
It is advisable to protect the router control plane by implementing
mechanisms to filter completely or rate-limit traffic not required at
the control plane level (i.e., unwanted traffic). "Router control
plane protection" is the concept of filtering or rate-limiting
unwanted traffic that would be diverted from the forwarding plane up
to the router control plane. The closer the filters and rate
limiters are to the forwarding plane and line-rate hardware, the more
effective the protection is and the more resistant the system is to
DoS attacks. This memo demonstrates one example of how to deploy a
policy filter that satisfies a set of sample traffic-matching,
filtering, and rate-limiting criteria.
Dugal, et al. Informational [Page 3]
RFC 6192 Protect Router Control Plane March 2011
Note that the filters described in this memo are applied only to
traffic that is destined for the router, and not to all traffic that
is passing through the router.
2. Applicability Statement
The method described in Section 3 and depicted in Figure 1
illustrates how to protect the router control plane from unwanted
traffic. Recognizing that deployment scenarios will vary, the exact
implementation is not generally applicable in all situations. The
categorization of legitimate router control plane traffic is
critically important in a successful implementation.
The examples given in this memo are simplified and minimalistic,
designed to illustrate the concept of protecting the router's control
plane. From them, operators can extrapolate specifics based on their
unique configuration and environment. This document is about
semantics, and Appendix A exemplifies syntax. For additional router
vendor implementations, or other converged devices, the syntax should
be translated to the respective language in a manner that preserves
the semantics.
Additionally, the need to provide the router control plane with
isolation, stability, and protection against rogue packets has been
incorporated into router designs for some time. Consequently, there
may be other vendor or implementation specific router control plane
protection mechanisms that are active by default or always active.
Those approaches may apply in conjunction with, or in addition to,
the method described in Section 3 and illustrated in Appendices A.1
and A.2. Those implementations should be considered as part of an
overall traffic management plan but are outside the scope of this
document.
This method is applicable for IPv4 as well as IPv6 address families,
and the legitimate traffic example in Section 3.1 provides examples
for both.
3. Method
In this memo, the authors demonstrate how a filter protecting the
router control plane can be deployed. In Section 3.1, a sample
router is introduced, and all traffic that its control plane must
process is identified. In Section 3.2, filter design concepts are
discussed. Cisco (Cisco IOS software) and Juniper (JUNOS)
implementations are provided in Appendices A.1 and A.2, respectively.
Dugal, et al. Informational [Page 4]
RFC 6192 Protect Router Control Plane March 2011
3.1. Legitimate Traffic
In this example, the router control plane must process traffic (i.e.,
traffic destined to the router and not through the router) per the
following criteria:
o Drop all IP packets that are fragments (see Section 3.3)
o Permit ICMP and ICMPv6 traffic from any source, rate-limited to
500 kbps for each category
o Permit OSPF traffic from routers within subnet 192.0.2.0/24 and
OSPFv3 traffic from IPv6 Link-Local unicast addresses (fe80::/10)
o Permit internal BGP (iBGP) traffic from routers within subnets
192.0.2.0/24 and 2001:db8:1::/48
o Permit external BGP (eBGP) traffic from eBGP peers 198.51.100.25,
198.51.100.27, 198.51.100.29, and 198.51.100.31; and IPv6 peers
2001:db8:100::25, 2001:db8:100::27, 2001:db8:100::29, and
2001:db8:100::31
o Permit DNS traffic from DNS servers within subnet 198.51.100.0/30
and 2001:db8:100:1::/64
o Permit NTP traffic from NTP servers within subnet 198.51.100.4/30
and 2001:db8:100:2::/64
o Permit Secure SHell (SSH) traffic from network management stations
within subnet 198.51.100.128/25 and 2001:db8:100:3::/64
o Permit Simple Network Management Protocol (SNMP) traffic from
network management stations within subnet 198.51.100.128/25 and
2001:db8:100:3::/64
o Permit RADIUS authentication and accounting replies from RADIUS
servers 198.51.100.9, 198.51.100.10, 2001:db8:100::9, and
2001:db8:100::10 that are listening on UDP ports 1812 and 1813
(Internet Assigned Numbers Authority (IANA) RADIUS ports). Note
that this does not accommodate a server using the original UDP
ports of 1645 and 1646
Dugal, et al. Informational [Page 5]
RFC 6192 Protect Router Control Plane March 2011
o Permit all other IPv4 and IPv6 traffic that was not explicitly
matched in a class above, rate-limited to 500 kbps, and drop above
that rate for each category
o Permit non-IP traffic (e.g., Connectionless Network Service
(CLNS), Internetwork Packet Exchange (IPX), PPP Link Control
Protocol (LCP), etc.), rate-limited to 250 kbps, and drop all
remaining traffic above that rate
The characteristics of legitimate traffic will vary from network to
network. To illustrate this, a router implementing the DHCP relay
function can rate-limit inbound DHCP traffic from clients and
restrict traffic from servers to a list of known DHCP servers. The
list of criteria above is provided for example only.
3.2. Filter Design
A filter is installed on the forwarding plane. This filter counts
and applies the actions to the categories of traffic described in
Section 3.1. Because the filter is enforced in the forwarding plane,
it prevents traffic from consuming bandwidth on the interface that
connects the forwarding plane to the router control plane. The
counters serve as an important forensic tool for the analysis of
potential attacks, and as an invaluable debugging and troubleshooting
aid. By adjusting the granularity and order of the filters, more
granular forensics can be performed (i.e., create a filter that
matches only traffic allowed from a group of IP addresses for a given
protocol followed by a filter that denies all traffic for that
protocol). This would allow for counters to be monitored for the
allowed protocol filter, as well as any traffic matching the specific
protocol that didn't originate from the explicitly allowed hosts.
In addition to the filters, rate limiters for certain classes of
traffic are also installed in the forwarding plane as defined in
Section 3.1. These rate limiters help further control the traffic
that will reach the router control plane for each filtered class as
well as all traffic not matching an explicit class. The actual rates
selected for various classes are network deployment specific;
analysis of the rates required for stability should be done
periodically. It is important to note that the most significant
factor to consider regarding the traffic profile going to the router
control plane is the packets per second (pps) rate. Therefore,
careful consideration must be given to determine the maximum pps rate
that could be generated from a given set of packet size and bandwidth
usage scenarios.
Dugal, et al. Informational [Page 6]
RFC 6192 Protect Router Control Plane March 2011
Syntactically, these filters explicitly define "allowed" traffic
(including IP addresses, protocols, and ports), define acceptable
actions for these acceptable traffic profiles (e.g., rate-limit or
simply permit the traffic), and then discard all traffic destined to
the router control plane that is not within the specifications of the
policy definition.
In an actual production environment, predicting a complete and
exhaustive list of traffic necessary to reach the router's control
plane for day-to-day operation may not be as obvious as the example
described herein. One recommended method to gauge this set of
traffic is to allow all traffic initially, and audit the traffic that
reaches the router control plane before applying any explicit filters
or rate limits. See Section 3.3 below for more discussion of this
topic.
The filter design provided in this document is intentionally limited
to attachment at the local router in question (e.g., a "service-
policy" attached to the "control-plane" in Cisco IOS, or a firewall
filter attached to the "lo0" interface in JUNOS). While virtually
all production environments utilize and rely heavily upon edge
protection or interface filtering, these methods of router protection
are beyond the intended scope of this document. Additionally, the
protocols themselves that are allowed to reach the router control
plane (e.g., OSPF, RSVP, TCP, SNMP, DNS, NTP, and inherently, SSH,
TLS, ESP, etc.) may have cryptographic security methods applied to
them, and the method of router control plane protection provided
herein is not a replacement for those cryptographic methods.
3.3. Design Trade-Offs
In designing the protection method, there are two independent parts
to consider: the classification of traffic (i.e., which traffic is
matched by the filters), and the policy actions taken on the
classified traffic (i.e., drop, permit, rate-limit, etc.).
There are different levels of granularity utilized for traffic
classification. For example, allowing all traffic from specific
source IP addresses versus allowing only a specific set of protocols
from those specific source IP addresses will each affect a different
subset of traffic.
Similarly, the policy actions taken on the classified traffic have
degrees of impact that may not become immediately obvious. For
example, discarding all ICMP traffic will have a negative impact on
the operational use of ICMP tools such as ping or traceroute to debug
network issues or to test deployment of a new circuit. Expanding on
this, in a real production network, an astute operator could define
Dugal, et al. Informational [Page 7]
RFC 6192 Protect Router Control Plane March 2011
varying rate limits for ICMP such that internal traffic is granted
uninhibited access to the router control plane, while traffic from
external addresses is rate-limited. Operators should pay special
attention to the new functionality and roles that ICMPv6 has in the
overall operation of IPv6 when designing the rate-limit policies.
Example functions include Neighbor Discovery (ND) and Multicast
Listener Discovery version 2 (MLDv2).
It is important to note that both classification and policy action
decisions are accompanied by respective trade-offs. Two examples of
these trade-off decisions are operational complexity at the expense
of policy and statistics-gathering detail, and tighter protection at
the expense of network supportability and troubleshooting ability.
Two types of traffic that need special consideration are IP fragments
and IP optioned packets:
o For network deployments where IP fragmentation is necessary, a
blanket policy of dropping all fragments destined to the router
control plane may not be feasible. However, many deployments
allow network configurations such that the router control plane
should never see a fragmented datagram. Since many attacks rely
on IP fragmentation, the example policy included herein drops all
fragments destined to the router control plane.
o Similarly, some deployments may choose to drop all IP optioned
packets. Others may need to loosen the constraint to allow for
protocols that require IP optioned packets such as the Resource
Reservation Protocol (RSVP). The design trade-off is that
dropping all IP optioned packets protects the router from attacks
that leverage malformed options, as well as attacks that rely on
the slow-path processing (i.e., software processing path) of IP
optioned packets. For network deployments where the protocols do
not use IP options, the filter is simpler to design in that it can
drop all packets with any IP option set. However, for networks
utilizing protocols relying on IP options, the filter to identify
the legitimate packets is more complex. If the filter is not
designed correctly, it could result in the inadvertent blackholing
of traffic for those protocols. This document does not include
filter configurations for IP optioned packets; additional
explanations regarding the filtering of packets based on the IP
options they contain can be found in [IP-OPTIONS-FILTER].
Dugal, et al. Informational [Page 8]
RFC 6192 Protect Router Control Plane March 2011
The goal of the method for protecting the router control plane is to
minimize the possibility for disruptions by reducing the vulnerable
surface, which is inversely proportional to the granularity of the
filter design. The finer the granularity of the filter design (e.g.,
filtering a more targeted subset of traffic from the rest of the
policed traffic, or isolating valid source addresses into a different
class or classes), the smaller the probability of disruption.
In addition to the traffic that matches explicit classes, care should
be taken on the policy decision that governs the handling of traffic
that would fall through the classification. Typically, that traffic
is referred to as traffic that gets matched in a default class. It
may also be traffic that matches a blanket protocol specific class
where previous classes that have more granular classification did not
match all packets for that specific protocol. The ideal policy would
have explicit classes to match only the traffic specifically required
at the router control plane and would drop all other traffic that
does not match a predefined class. As most vendor implementations
permit all traffic hitting the default class, an explicit drop action
would need to be configured in the policy such that the traffic
hitting that default class would be dropped, versus being permitted
and delivered to the router control plane. This approach requires
rigorous traffic pattern identification such that a default drop
policy does not break existing device functionality. The approach
defined in this document allows the default traffic and rate-limits
it as opposed to dropping it. This approach was chosen as a way to
give the operator time to evaluate and characterize traffic in a
production scenario prior to dropping all traffic not explicitly
matched and permitted. However, it is highly recommended that after
monitoring the traffic matching the default class, explicit classes
be defined to catch the legitimate traffic. After all legitimate
traffic has been identified and explicitly allowed, the default class
should be configured to drop any remaining traffic.
Additionally, the baselining and monitoring of traffic flows to the
router's control plane are critical in determining both the rates and
granularity of the policies being applied. It is also important to
validate the existing policies and rules or update them as the
network evolves and its traffic dynamics change. Some possible ways
to achieve this include individual policy counters that can be
exported or retrieved, for example via SNMP, and logging of filtering
actions.
Finally, the use of flow-based behavioral analysis or command-line
interface (CLI) functions to identify what client/server functions a
given router's control plane handles would be very useful during
initial policy development phases, and certainly for ongoing forensic
analysis.
Dugal, et al. Informational [Page 9]
RFC 6192 Protect Router Control Plane March 2011
3.4. Additional Protection Considerations
In addition to the design described in this document of defining
"allowed" traffic (i.e., identifying traffic that the control plane
must process) and limiting (e.g., rate-limiting or blocking) the
rest, the router control plane protection method can be applied to
thwart specific attacks. In particular, it can be used to protect
against TCP SYN flooding attacks and other Denial-of-Service attacks
that starve router control plane resources.
4. Security Considerations
The filters described in this document leave the router susceptible
to discovery from any host in the Internet. If network operators
find this risk objectionable, they can reduce the exposure to
discovery with ICMP by restricting the sub-networks from which ICMP
Echo requests and potential traceroute packets (i.e., packets that
would trigger an ICMP Time Exceeded reply) are accepted, and
therefore to which sub-networks ICMP responses (ICMP Echo Reply and
Time Exceeded) are sent. A similar concern exists for ICMPv6 traffic
but on a broader level due to the additional functionalities
implemented in ICMPv6. Filtering recommendations for ICMPv6 can be
found in [RFC4890]. Moreover, different rate-limiting policies may
be defined for internally (e.g., from the Network Operations Center
(NOC)) versus externally sourced traffic. Note that this document is
not targeted at the specifics of ICMP filtering or traffic filtering
designed to prevent device discovery.
The filters described in this document do not block unwanted traffic
having spoofed source addresses that match a defined traffic profile
as discussed in Section 3.1. Network operators can mitigate this
risk by preventing source address spoofing with filters applied at
the network edge. Refer to Section 5.3.8 of [RFC1812] for more
information regarding source address validation. Other methods also
exist for limiting exposure to packet spoofing, such as the
Generalized Time to Live (TTL) Security Mechanism (GTSM) [RFC5082]
and Ingress Filtering [RFC2827] [RFC3704].
The ICMP rate limiter specified for the filters described in this
document protects the router from floods of ICMP traffic; see
Sections 3.1 and 3.3 for details. However, during an ICMP flood,
some legitimate ICMP traffic may be dropped. Because of this, when
operators discover a flood of ICMP traffic, they are highly motivated
to stop it at the source where the traffic is being originated.
Dugal, et al. Informational [Page 10]
RFC 6192 Protect Router Control Plane March 2011
Additional considerations pertaining to the usage and handling of
traffic that utilizes the IP Router Alert Options can be found in
[RTR-ALERT-CONS], and additional IP options filtering explanations
can be found in [IP-OPTIONS-FILTER].
The treatment of exception traffic in the forwarding plane and the
generation of specific messages by the router control plane also
require protection from a DoS attack. Specifically, the generation
of ICMP Unreachable messages by the router control plane needs to be
rate-limited, either implicitly within the router's architecture or
explicitly through configuration. When possible, different ICMP
Destination Unreachable codes (e.g., "fragmentation needed and DF
set") or "Packet Too Big" messages can receive a different rate-
limiting treatment. Continuous benchmarking of router-generated ICMP
traffic should be done before applying rate limits such that
sufficient headroom is included to prevent inadvertent Path Maximum
Transmission Unit Discovery (PMTUD) blackhole scenarios during normal
operation. It is also recommended to deploy explicit rate limiters
where possible to improve troubleshooting and monitoring capability.
The explicit rate limiters in a class allow for monitoring tools to
detect and report when these rate limiters become active (i.e., when
traffic is policed). This in turn serves as an indicator that either
the normal traffic rates have increased or "out of policy" traffic
rates have been detected. More thorough analysis of the traffic
flows and rate-limited traffic is needed to identify which of these
two cases triggered the rate limiters. For additional information
regarding specific ICMP rate-limiting, see Section 4.3.2.8 of
[RFC1812].
Additionally, the handling of TTL / Hop Limit expired traffic needs
protection. This traffic is not necessarily addressed to the device,
but it can get sent to the router control plane to process the TTL /
Hop Limit expiration. For example, rate-limiting the TTL / Hop Limit
expired traffic before sending the packets to the router control
plane component that will generate the ICMP error, and distributing
the sending of ICMP errors to Line Card CPUs, are protection
mechanisms that mitigate attacks before they can negatively affect a
rate-limited router control plane component.
5. Acknowledgements
The authors would like to thank Ron Bonica for providing initial and
ongoing review, suggestions, and valuable input. Pekka Savola,
Warren Kumari, and Xu Chen provided very thorough and useful feedback
that improved the document. Many thanks to John Kristoff,
Christopher Morrow, and Donald Smith for a fruitful discussion around
the operational and manageability aspects of router control plane
protection techniques. The authors would also like to thank
Dugal, et al. Informational [Page 11]
RFC 6192 Protect Router Control Plane March 2011
Joel Jaeggli, Richard Graveman, Danny McPherson, Gregg Schudel, Eddie
Parra, Seo Boon Ng, Manav Bhatia, German Martinez, Wen Zhang, Roni
Even, Acee Lindem, Glen Zorn, Joe Abley, Ralph Droms, and Stewart
Bryant for providing thorough review, useful suggestions, and
valuable input. Assistance from Jim Bailey and Raphan Han in
providing technical direction and sample configuration guidance on
the IPv6 sections was also very much appreciated. Finally, the
authors extend kudos to Andrew Yourtchenko for his review, comments,
and willingness to present this document at IETF 78 (July 2010,
Maastricht, The Netherlands) on behalf of the authors.
6. Informative References
[IP-OPTIONS-FILTER]
Gont, F. and S. Fouant, "IP Options Filtering
Recommendations", Work in Progress, February 2010.
[RFC1812] Baker, F., Ed., "Requirements for IP Version 4 Routers",
RFC 1812, June 1995.
[RFC2827] Ferguson, P. and D. Senie, "Network Ingress Filtering:
Defeating Denial of Service Attacks which employ IP Source
Address Spoofing", BCP 38, RFC 2827, May 2000.
[RFC3704] Baker, F. and P. Savola, "Ingress Filtering for Multihomed
Networks", BCP 84, RFC 3704, March 2004.
[RFC4890] Davies, E. and J. Mohacsi, "Recommendations for Filtering
ICMPv6 Messages in Firewalls", RFC 4890, May 2007.
[RFC5082] Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C.
Pignataro, "The Generalized TTL Security Mechanism
(GTSM)", RFC 5082, October 2007.
[RTR-ALERT-CONS]
Le Faucheur, F., Ed., "IP Router Alert Considerations and
Usage", Work in Progress, March 2011.
Dugal, et al. Informational [Page 12]
RFC 6192 Protect Router Control Plane March 2011
Appendix A. Configuration Examples
The configurations provided below are syntactical representations of
the semantics described in the document and should be treated as
non-normative.
A.1. Cisco Configuration
Refer to the Control Plane Policing (CoPP) document in the Cisco IOS
Software Feature Guides (available at
RFC, FYI, BCP