|
Network Working Group Request for Comments: 4552 Category: Standards Track |
M. Gupta Tropos Networks N. Melam Juniper Networks June 2006 |
This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.
Copyright © The Internet Society (2006).
This document describes means and mechanisms to provide
authentication/confidentiality to OSPFv3 using an IPv6 Authentication
Header/Encapsulating Security Payload (AH/ESP) extension header.
1. Introduction
1.1. Conventions Used in This Document
2. Transport Mode vs. Tunnel Mode
3. Authentication
4. Confidentiality
5. Distinguishing OSPFv3 from OSPFv2
6. IPsec Requirements
7. Key Management
8. SA Granularity and Selectors
9. Virtual Links
10. Rekeying
10.1. Rekeying Procedure
10.2. KeyRolloverInterval
10.3. Rekeying Interval
11. IPsec Protection Barrier and SPD
12. Entropy of Manual Keys
13. Replay Protection
14. Security Considerations
15. References
15.1. Normative References
15.2. Informative References
OSPF (Open Shortest Path First) Version 2 [N1] defines the fields AuType and Authentication in its protocol header to provide security. In OSPF for IPv6 (OSPFv3) [N2], both of the authentication fields were removed from OSPF headers. OSPFv3 relies on the IPv6 Authentication Header (AH) and IPv6 Encapsulating Security Payload (ESP) to provide integrity, authentication, and/or confidentiality.
This document describes how IPv6 AH/ESP extension headers can be used to provide authentication/confidentiality to OSPFv3.
It is assumed that the reader is familiar with OSPFv3 [N2], AH [N5], ESP [N4], the concept of security associations, tunnel and transport mode of IPsec, and the key management options available for AH and ESP (manual keying [N3] and Internet Key Exchange (IKE)[I1]).
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 [N7].
The transport mode Security Association (SA) is generally used between two hosts or routers/gateways when they are acting as hosts. The SA must be a tunnel mode SA if either end of the security association is a router/gateway. Two hosts MAY establish a tunnel mode SA between themselves. OSPFv3 packets are exchanged between routers. However, since the packets are locally delivered, the routers assume the role of hosts in the context of tunnel mode SA. All implementations conforming to this specification MUST support transport mode SA to provide required IPsec security to OSPFv3 packets. They MAY also support tunnel mode SA to provide required IPsec security to OSPFv3 packets.
Implementations conforming to this specification MUST support authentication for OSPFv3.
In order to provide authentication to OSPFv3, implementations MUST support ESP and MAY support AH.
If ESP in transport mode is used, it will only provide authentication to OSPFv3 protocol packets excluding the IPv6 header, extension headers, and options.
If AH in transport mode is used, it will provide authentication to OSPFv3 protocol packets, selected portions of IPv6 header, selected portions of extension headers, and selected options.
When OSPFv3 authentication is enabled,
Implementations conforming to this specification SHOULD support confidentiality for OSPFv3.
If confidentiality is provided, ESP MUST be used.
When OSPFv3 confidentiality is enabled,
The IP/IPv6 Protocol Type for OSPFv2 and OSPFv3 is the same (89), and OSPF distinguishes them based on the OSPF header version number. However, current IPsec standards do not allow using arbitrary protocol-specific header fields as the selectors. Therefore, the OSPF version field in the OSPF header cannot be used to distinguish OSPFv3 packets from OSPFv2 packets. As OSPFv2 is only for IPv4 and OSPFv3 is only for IPv6, the version field in the IP header can be used to distinguish OSPFv3 packets from OSPFv2 packets.
In order to implement this specification, the following IPsec capabilities are required.
Transport mode
IPsec in transport mode MUST be supported. [N3]
Multiple Security Policy Databases (SPDs)
The implementation MUST support multiple SPDs with an SPD
selection function that provides an ability to choose a specific
SPD based on interface. [N3]
Selectors
The implementation MUST be able to use source address, destination
address, protocol, and direction as selectors in the SPD.
Interface ID tagging
The implementation MUST be able to tag the inbound packets with
the ID of the interface (physical or virtual) via which it
arrived. [N3]
Manual key support
Manually configured keys MUST be able to secure the specified
traffic. [N3]
Encryption and authentication algorithms
The implementation MUST NOT allow the user to choose stream
ciphers as the encryption algorithm for securing OSPFv3 packets
since the stream ciphers are not suitable for manual keys.
Except when in conflict with the above statement, the key words "MUST", "MUST NOT", "REQUIRED", "SHOULD", and "SHOULD NOT" that appear in the [N6] document for algorithms to be supported are to be interpreted as described in [N7] for OSPFv3 support as well.
Dynamic IPsec rule configuration
The routing module SHOULD be able to configure, modify, and delete
IPsec rules on the fly. This is needed mainly for securing
virtual links.
Encapsulation of ESP packet
IP encapsulation of ESP packets MUST be supported. For
simplicity, UDP encapsulation of ESP packets SHOULD NOT be used.
Different SAs for different Differentiated Services Code Points
(DSCPs)
As per [N3], the IPsec implementation MUST support the
establishment and maintenance of multiple SAs with the same
selectors between a given sender and receiver. This allows the
implementation to associate different classes of traffic with the
same selector values in support of Quality of Service (QoS).
OSPFv3 exchanges both multicast and unicast packets. While running OSPFv3 over a broadcast interface, the authentication/confidentiality required is "one to many". Since IKE is based on the Diffie-Hellman key agreement protocol and works only for two communicating parties, it is not possible to use IKE for providing the required "one to many" authentication/confidentiality. This specification mandates the usage of Manual Keying with current IPsec implementations. Future specifications can explore the usage of protocols like Kerberized Internet Negotiation of Keys/Group Secure Association Key Management Protocol (KINK/GSAKMP) when they are widely available. In manual keying, SAs are statically installed on the routers and these static SAs are used to authenticate/encrypt packets.
The following discussion explains that it is not scalable and is practically infeasible to use different security associations for inbound and outbound traffic to provide the required "one to many" security. Therefore, the implementations MUST use manually
configured keys with the same SA parameters (Security Parameter Index (SPI), keys, etc.) for both inbound and outbound SAs (as shown in Figure 3).
A |
SAa ------------>|
SAb <------------|
|
B |
SAb ------------>|
SAa <------------| Figure 1
|
C |
SAa/SAb ------------>|
SAa/SAb <------------|
|
Broadcast
Network
If we consider communication between A and B in Figure 1, everything seems to be fine. A uses security association SAa for outbound packets and B uses the same for inbound packets and vice versa. Now if we include C in the group and C sends a packet using SAa, then only A will be able to understand it. Similarly, if C sends a packet using SAb, then only B will be able to understand it. Since the packets are multicast and they are going to be processed by both A and B, there is no SA for C to use so that both A and B can understand them.
A |
SAa ------------>|
SAb <------------|
SAc <------------|
|
B |
SAb ------------>|
SAa <------------| Figure 2
SAc <------------|
|
C |
SAc ------------>|
SAa <------------|
SAb <------------|
|
Broadcast
Network
The problem can be solved by configuring SAs for all the nodes on every other node as shown in Figure 2. So A, B, and C will use SAa, SAb, and SAc, respectively, for outbound traffic. Each node will lookup the SA to be used based on the source (A will use SAb and SAc for packets received from B and C, respectively). This solution is not scalable and practically infeasible because a large number of SAs will need to be configured on each node. Also, the addition of a node in the broadcast network will require the addition of another SA on every other node.
A |
SAo ------------>|
SAi <------------|
|
B |
SAo ------------>|
SAi <------------| Figure 3
|
C |
SAo ------------>|
SAi <------------|
|
Broadcast
Network
The problem can be solved by using the same SA parameters (SPI, keys, etc.) for both inbound (SAi) and outbound (SAo) SAs as shown in Figure 3.
The user SHOULD be given the choice of sharing the same SA among multiple interfaces or using a unique SA per interface.
OSPFv3 supports running multiple instances over one interface using the "Instance Id" field contained in the OSPFv3 header. As IPsec does not support arbitrary fields in the protocol header to be used as the selectors, it is not possible to use different SAs for different OSPFv3 instances running over the same interface. Therefore, all OSPFv3 instances running over the same interface will have to use the same SA. In OSPFv3 RFC terminology, SAs are per-link and not per-interface.
A different SA than the SA of the underlying interface MUST be provided for virtual links. Packets sent on virtual links use unicast non-link local IPv6 addresses as the IPv6 source address, while packets sent on other interfaces use multicast and unicast link local addresses. This difference in the IPv6 source address differentiates the packets sent on virtual links from other OSPFv3 interface types.
As the virtual link end point IPv6 addresses are not known, it is not possible to install SPD/Security Association Database (SAD) entries at the time of configuration. The virtual link end point IPv6 addresses are learned during the routing table computation process. The packet exchange over the virtual links starts only after the discovery of the end point IPv6 addresses. In order to protect these exchanges, the routing module must install the corresponding SPD/SAD entries before starting these exchanges. Note that manual SA parameters are preconfigured but not installed in the SAD until the end point addresses are learned.
According to the OSPFv3 RFC [N2], the virtual neighbor's IP address is set to the first prefix with the "LA-bit" set from the list of prefixes in intra-area-prefix-Link State Advertisements (LSAs) originated by the virtual neighbor. But when it comes to choosing the source address for the packets that are sent over the virtual link, the RFC [N2] simply suggests using one of the router's own global IPv6 addresses. In order to install the required security rules for virtual links, the source address also needs to be predictable. Hence, routers that implement this specification MUST change the way the source and destination addresses are chosen for packets exchanged over virtual links when IPsec is enabled.
The first IPv6 address with the "LA-bit" set in the list of prefixes advertised in intra-area-prefix-LSAs in the transit area MUST be used as the source address for packets exchanged over the virtual link. When multiple intra-area-prefix-LSAs are originated, they are considered concatenated and are ordered by ascending Link State ID.
The first IPv6 address with the "LA-bit" set in the list of prefixes received in intra-area-prefix-LSAs from the virtual neighbor in the transit area MUST be used as the destination address for packets exchanged over the virtual link. When multiple intra-area-prefix- LSAs are received, they are considered concatenated and are ordered by ascending Link State ID.
This makes both the source and destination addresses of packets exchanged over the virtual link predictable when IPsec is enabled.
To maintain the security of a link, the authentication and encryption key values SHOULD be changed periodically.
The following three-step procedure SHOULD be provided to rekey the routers on a link without dropping OSPFv3 protocol packets or disrupting the adjacency.
(1) For every router on the link, create an additional inbound SA for the interface being rekeyed using a new SPI and the new key.
(2) For every router on the link, replace the original outbound SA with one using the new SPI and key values. The SA replacement operation should be atomic with respect to sending OSPFv3 packets on the link so that no OSPFv3 packets are sent without authentication/encryption.
(3) For every router on the link, remove the original inbound SA.
Note that all routers on the link must complete step 1 before any begin step 2. Likewise, all the routers on the link must complete step 2 before any begin step 3.
One way to control the progression from one step to the next is for each router to have a configurable time constant KeyRolloverInterval. After the router begins step 1 on a given link, it waits for this interval and then moves to step 2. Likewise, after moving to step 2, it waits for this interval and then moves to step 3.
In order to achieve smooth key transition, all routers on a link should use the same value for KeyRolloverInterval and should initiate the key rollover process within this time period.
At the end of this procedure, all the routers on the link will have a single inbound and outbound SA for OSPFv3 with the new SPI and key values.
The configured value of KeyRolloverInterval should be long enough to allow the administrator to change keys on all the OSPFv3 routers. As this value can vary significantly depending upon the implementation and the deployment, it is left to the administrator to choose an appropriate value.
This section analyzes the security provided by manual keying and recommends that the encryption and authentication keys SHOULD be changed at least every 90 days.
The weakest security provided by the security mechanisms discussed in this specification is when NULL encryption (for ESP) or no encryption (for AH) is used with the HMAC-MD5 authentication. Any other algorithm combinations will at least be as hard to break as the ones mentioned above. This is shown by the following reasonable assumptions:
RFC 3562 [I4] analyzes the rekeying requirements for the TCP MD5 signature option. The analysis provided in RFC 3562 is also applicable to this specification as the analysis is independent of data patterns.
The IPsec protection barrier MUST be around the OSPF protocol. Therefore, all the inbound and outbound OSPF traffic goes through IPsec processing.
The SPD selection function MUST return an SPD with the following rule
for all the interfaces that have OSPFv3
authentication/confidentiality disabled.
No. source destination protocol action
1 any any OSPF bypass
The SPD selection function MUST return an SPD with the following
rules for all the interfaces that have OSPFv3
authentication/confidentiality enabled.
No. source destination protocol action
2 fe80::/10 any OSPF protect
3 fe80::/10 any ESP/OSPF or AH/OSPF protect
4 src/128 dst/128 OSPF protect
5 src/128 dst/128 ESP/OSPF or AH/OSPF protect
For rules 2 and 4, action "protect" means encrypting/calculating Integrity Check Value (ICV) and adding an ESP or AH header. For rules 3 and 5, action "protect" means decrypting/authenticating the packets and stripping the ESP or AH header.
Rule 1 will bypass the OSPFv3 packets without any IPsec processing on the interfaces that have OSPFv3 authentication/confidentiality disabled.
Rules 2 and 4 will drop the inbound OSPFv3 packets that have not been secured with ESP/AH headers.
ESP/OSPF or AH/OSPF in rules 3 and 5 mean that it is an OSPF packet secured with ESP or AH.
Rules 2 and 3 are meant to secure the unicast and multicast OSPF packets that are not being exchanged over the virtual links.
Rules 4 and 5 are meant to secure the packets being exchanged over virtual links. These rules are installed after learning the virtual link end point IPv6 addresses. These rules MUST be installed in the SPD for the interfaces that are connected to the transit area for the virtual link. These rules MAY alternatively be installed on all the interfaces. If these rules are not installed on all the interfaces, clear text or malicious OSPFv3 packets with the same source and destination addresses as the virtual link end point IPv6 addresses will be delivered to OSPFv3. Though OSPFv3 drops these packets as they were not received on the right interface, OSPFv3 receives some clear text or malicious packets even when the security is enabled. Installing these rules on all the interfaces ensures that OSPFv3 does not receive these clear text or malicious packets when security is enabled. On the other hand, installing these rules on all the interfaces increases the processing overhead on the interfaces where there is no other IPsec processing. The decision of whether to install these rules on all the interfaces or on just the interfaces that are connected to the transit area is a private decision and doesn't affect the interoperability in any way. Hence it is an implementation choice.
The implementations MUST allow the administrator to configure the cryptographic and authentication keys in hexadecimal format rather than restricting it to a subset of ASCII characters (letters, numbers, etc.). A restricted character set will reduce key entropy significantly as discussed in [I2].
Since it is not possible using the current standards to provide complete replay protection while using manual keying, the proposed solution will not provide protection against replay attacks.
Detailed analysis of various vulnerabilities of the routing protocols and OSPF in particular is discussed in [I3] and [I2]. The conclusion is that replay of OSPF packets can cause adjacencies to be disrupted, which can lead to a DoS attack on the network. It can also cause database exchange process to occur continuously thus causing CPU overload as well as micro loops in the network.
This memo discusses the use of IPsec AH and ESP headers to provide security to OSPFv3 for IPv6. Hence, security permeates throughout this document.
OSPF Security Vulnerabilities Analysis [I2] identifies OSPF vulnerabilities in two scenarios -- one with no authentication or simple password authentication and the other with cryptographic authentication. The solution described in this specification provides protection against all the vulnerabilities identified for scenarios with cryptographic authentication with the following exceptions:
Limitations of manual key:
This specification mandates the usage of manual keys. The following are the known limitations of the usage of manual keys.
Impersonating attacks:
The usage of the same key on all the OSPF routers connected to a link leaves them all insecure against impersonating attacks if any one of the OSPF routers is compromised, malfunctioning, or misconfigured.
Detailed analysis of various vulnerabilities of routing protocols is discussed in [I3].
[N1] Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998.
[N2] Coltun, R., Ferguson, D., and J. Moy, "OSPF for IPv6", RFC 2740, December 1999.
[N3] Kent, S. and K. Seo, "Security Architecture for the Internet Protocol", RFC 4301, December 2005.
[N4] Kent, S., "IP Encapsulating Security Payload (ESP)", RFC 4303, December 2005.
[N5] Kent, S., "IP Authentication Header", RFC 4302, December 2005.
[N6] Eastlake 3rd, D., "Cryptographic Algorithm Implementation Requirements for Encapsulating Security Payload (ESP) and Authentication Header (AH)", RFC 4305, December 2005.
[N7] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997.
[I1] Kaufman, C., "Internet Key Exchange (IKEv2) Protocol", RFC 4306, December 2005.
[I2] Jones, E. and O. Moigne, "OSPF Security Vulnerabilities Analysis", Work in Progress.
[I3] Barbir, A., Murphy, S., and Y. Yang, "Generic Threats to Routing Protocols", Work in Progress.
[I4] Leech, M., "Key Management Considerations for the TCP MD5 Signature Option", RFC 3562, July 2003.
The authors would like to extend sincere thanks to Marc Solsona, Janne Peltonen, John Cruz, Dhaval Shah, Abhay Roy, Paul Wells, Vishwas Manral, and Sam Hartman for providing useful information and critiques on this memo. The authors would like to extend special thanks to Acee Lindem for many editorial changes.
We would also like to thank members of the IPsec and OSPF WG for providing valuable review comments.
Mukesh Gupta
Tropos Networks
555 Del Rey Ave
Sunnyvale, CA 94085
Phone: 408-331-6889
EMail: mukesh.gupta@tropos.com
Nagavenkata Suresh Melam
Juniper Networks
1194 N. Mathilda Ave
Sunnyvale, CA 94089
Phone: 408-505-4392
EMail: nmelam@juniper.net
Copyright © The Internet Society (2006).
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.
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.
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.
Funding for the RFC Editor function is provided by the IETF Administrative Support Activity (IASA).