]> RFG Security, LLC

15 Park Avenue Morristown NJ 07960 USA rfg@acm.org

Nokia

313 Fairchild Drive Mountain View CA 94043 USA mohanp@sbcglobal.net

CSC/FUNET

Espoo Finland psavola@funet.fi

Nokia Siemens Network

Otto-Hahn-Ring 6 Munich Bayern 81739 Germany Hannes.Tschofenig@nsn.com

Operations & Management IPv6 Operations WG This document gives guidance on securing manually configured IPv6-in-IPv4 tunnels using IPsec in transport mode. No additional protocol extensions are described beyond those available with the IPsec framework.

The IPv6 Operations (v6ops) working group has selected (manually configured) IPv6-in-IPv4 tunneling as one of the IPv6 transition mechanisms for IPv6 deployment. identified a number of threats that had not been adequately analyzed or addressed in its predecessor . The most complete solution is to use IPsec to protect IPv6-in-IPv4 tunneling. The document was intentionally not expanded to include the details on how to set up an IPsec-protected tunnel in an interoperable manner, but instead the details were deferred to this memo. The first four sections of this document analyze the threats and scenarios that can be addressed by IPsec and assumptions made by this document for successful IPsec Security Association (SA) establishment. gives the details of Internet Key Exchange (IKE) and IP security (IPsec) exchange with packet formats and Security Policy Database (SPD) entries. gives recommendations. Appendices further discuss tunnel mode usage and optional extensions. This document does not address the use of IPsec for tunnels that are not manually configured (e.g., 6to4 tunnels ). Presumably, some form of opportunistic encryption or "better-than-nothing security" might or might not be applicable. Similarly, propagating quality-of-service attributes (apart from Explicit Congestion Notification bits ) from the encapsulated packets to the tunnel path is out of scope. The use of the word "interface" or the phrase "IP interface" refers to the IPv6 interface that must be present on any IPv6 node to send or receive IPv6 packets. The use of the phrase "tunnel interface" refers to the interface that receives the IPv6-in-IPv4 tunneled packets over IPv4.

is mostly concerned about address spoofing threats: The IPv4 source address of the encapsulating ("outer") packet can be spoofed. The IPv6 source address of the encapsulated ("inner") packet can be spoofed. The reason threat (1) exists is the lack of universal deployment of IPv4 ingress filtering . The reason threat (2) exists is that the IPv6 packet is encapsulated in IPv4 and hence may escape IPv6 ingress filtering. specifies the following strict address checks as mitigating measures: To mitigate threat (1), the decapsulator verifies that the IPv4 source address of the packet is the same as the address of the configured tunnel endpoint. The decapsulator may also implement IPv4 ingress filtering, i.e., check whether the packet is received on a legitimate interface. To mitigate threat (2), the decapsulator verifies whether the inner IPv6 address is a valid IPv6 address and also applies IPv6 ingress filtering before accepting the IPv6 packet. This memo proposes using IPsec for providing stronger security in preventing these threats and additionally providing integrity, confidentiality, replay protection, and origin protection between tunnel endpoints. IPsec can be used in two ways, in transport and tunnel mode; detailed discussion about applicability in this context is provided in .

In transport mode, the IPsec Encapsulating Security Payload (ESP) or Authentication Header (AH) security association (SA) is established to protect the traffic defined by (IPv4-source, IPv4-dest, protocol = 41). On receiving such an IPsec packet, the receiver first applies the IPsec transform (e.g., ESP) and then matches the packet against the Security Parameter Index (SPI) and the inbound selectors associated with the SA to verify that the packet is appropriate for the SA via which it was received. A successful verification implies that the packet came from the right IPv4 endpoint, because the SA is bound to the IPv4 source address. This prevents threat (1) but not threat (2). IPsec in transport mode does not verify the contents of the payload itself where the IPv6 addresses are carried. That is, two nodes using IPsec transport mode to secure the tunnel can spoof the inner payload. The packet will be decapsulated successfully and accepted. This shortcoming can be partially mitigated by IPv6 ingress filtering, i.e., check that the packet is arriving from the interface in the direction of the route towards the tunnel endpoint, similar to a Strict Reverse Path Forwarding (RPF) check . In most implementations, a transport mode SA is applied to a normal IPv6-in-IPv4 tunnel. Therefore, ingress filtering can be applied in the tunnel interface. (Transport mode is often also used in other kinds of tunnels such as Generic Routing Encapsulation (GRE) and Layer 2 Tunneling Protocol (L2TP) .)

In tunnel mode, the IPsec SA is established to protect the traffic defined by (IPv6-source, IPv6-destination). On receiving such an IPsec packet, the receiver first applies the IPsec transform (e.g., ESP) and then matches the packet against the SPI and the inbound selectors associated with the SA to verify that the packet is appropriate for the SA via which it was received. The successful verification implies that the packet came from the right endpoint. The outer IPv4 addresses may be spoofed, and IPsec cannot detect this in tunnel mode; the packets will be demultiplexed based on the SPI and possibly the IPv6 address bound to the SA. Thus, the outer address spoofing is irrelevant as long as the decryption succeeds and the inner IPv6 packet can be verified to have come from the right tunnel endpoint. As described in , using tunnel mode is more difficult than applying transport mode to a tunnel interface, and as a result this document recommends transport mode. Note that even though transport rather than tunnel mode is recommended, an IPv6-in-IPv4 tunnel specified by protocol 41 still exists .

There are roughly three scenarios: (Generic) router-to-router tunnels. Site-to-router or router-to-site tunnels. These refer to tunnels between a site's IPv6 (border) device and an IPv6 upstream provider's router. A degenerate case of a site is a single host. Host-to-host tunnels.

IPv6/IPv4 hosts and routers can tunnel IPv6 datagrams over regions of IPv4 forwarding topology by encapsulating them within IPv4 packets. Tunneling can be used in a variety of ways.

| Router | | A | (_ _) | B | '--------' '----' '--------' ^ IPsec tunnel between ^ | Router A and Router B | V V ]]>

IPv6/IPv4 routers interconnected by an IPv4 infrastructure can tunnel IPv6 packets between themselves. In this case, the tunnel spans one segment of the end-to-end path that the IPv6 packet takes. The source and destination addresses of the IPv6 packets traversing the tunnel could come from a wide range of IPv6 prefixes, so binding IPv6 addresses to be used to the SA is not generally feasible. IPv6 ingress filtering must be performed to mitigate the IPv6 address spoofing threat. A specific case of router-to-router tunnels, when one router resides at an end site, is described in the next section.

This is a generalization of host-to-router and router-to-host tunneling, because the issues when connecting a whole site (using a router) and connecting a single host are roughly equal.

| Router |<=======( Internet )=======>| Site B | (_ _) | A | (_ _) '--------' '----' '---------' '----' ^ | V .--------. | Native | | IPv6 | | node | '--------' ]]>

IPv6/IPv4 routers can tunnel IPv6 packets to their final destination IPv6/IPv4 site. This tunnel spans only the last segment of the end-to-end path.

| Router | | '--------' (_ _) | | A | | '----' | '--------' | IPsec tunnel between | ^ | IPv6 Site and Router A | | | | V | | .-------. | | | V6 | | | | Hosts | | | '--------' | +---------------------+ ]]>

--> In the other direction, IPv6/IPv4 hosts can tunnel IPv6 packets to an intermediary IPv6/IPv4 router that is reachable via an IPv4 infrastructure. This type of tunnel spans the first segment of the packet's end-to-end path. The hosts in the site originate the packets with IPv6 source addresses coming from a well-known prefix, whereas the destination addresses could be any nodes on the Internet. In this case, an IPsec tunnel mode SA could be bound to the prefix that was allocated to the router at Site B, and Router A could verify that the source address of the packet matches the prefix. Site B will not be able to do a similar verification for the packets it receives. This may be quite reasonable for most of the deployment cases, for example, an Internet Service Provider (ISP) allocating a /48 to a customer. The Customer Premises Equipment (CPE) where the tunnel is terminated "trusts" (in a weak sense) the ISP's router, and the ISP's router can verify that Site B is the only one that can originate packets within the /48. IPv6 spoofing must be prevented, and setting up ingress filtering may require some amount of manual configuration; see more of these options in .

| Host | | A | (_ _) | B | '--------' '----' '--------' IPsec tunnel between Host A and Host B ]]>

IPv6/IPv4 hosts interconnected by an IPv4 infrastructure can tunnel IPv6 packets between themselves. In this case, the tunnel spans the entire end-to-end path. In this case, the source and the destination IPv6 addresses are known a priori. A tunnel mode SA could be bound to these specific addresses. Address verification prevents IPv6 source address spoofing completely. As noted in the Introduction, automatic host-to-host tunneling methods (e.g., 6to4) are out of scope for this memo.

This section discusses the different versions of the IKE and IPsec security architecture and their applicability to this document. The IPsec security architecture was previously defined in and is now superseded by . IKE was originally defined in (which is called IKEv1 in this document) and is now superseded by (called IKEv2; see also ). There are several differences between them. The differences relevant to this document are discussed below. does not require allowing IP as the next layer protocol in traffic selectors when an IPsec SA is negotiated. In contrast, requires supporting IP as the next layer protocol (like TCP or UDP) in traffic selectors. assumes IKEv2, as some of the new features cannot be negotiated using IKEv1. It is valid to negotiate multiple traffic selectors for a given IPsec SA in . This is possible only with IKEv2. If IKEv1 is used, then multiple SAs need to be set up, one for each traffic selector. Note that the existing implementations based on IKEv1 may already be able to support the features described in (1) and (2). If appropriate, the deployment may choose to use either version of the security architecture. IKEv2 supports features useful for configuring and securing tunnels not present with IKEv1. IKEv2 supports legacy authentication methods by carrying them in Extensible Authentication Protocol (EAP) payloads. This can be used to authenticate hosts or sites to an ISP using EAP methods that support username and password. IKEv2 supports dynamic address configuration, which may be used to configure the IPv6 address of the host. Network Address Translation (NAT) traversal works with both the old and revised IPsec architectures, but the negotiation is integrated with IKEv2. For the purposes of this document, where the confidentiality of ESP is not required, AH can be used as an alternative to ESP. The main difference is that AH is able to provide integrity protection for certain fields in the outer IPv4 header and IPv4 options. However, as the outer IPv4 header will be discarded in any case, and those particular fields are not believed to be relevant in this particular application, there is no particular reason to use AH.

This section describes the SPD entries for setting up the IPsec transport mode SA to protect the IPv6 traffic. Several requirements arise when IPsec is used to protect the IPv6 traffic (inner header) for the scenarios listed in . All of IPv6 traffic should be protected, including link-local (e.g., Neighbor Discovery) and multicast traffic. Without this, an attacker can pollute the IPv6 neighbor cache causing disruption in communication between the two routers. In router-to-router tunnels, the source and destination addresses of the traffic could come from a wide range of prefixes that are normally learned through routing. As routing can always learn a new prefix, one cannot assume that all the prefixes are known a priori . This mainly affects scenario (1). Source address selection depends on the notions of routes and interfaces. This implies that the reachability to the various IPv6 destinations appear as routes in the routing table. This affects scenarios (2) and (3). The IPv6 traffic can be protected using transport or tunnel mode. There are many problems when using tunnel mode as implementations may or may not model the IPsec tunnel mode SA as an interface as described in . If IPsec tunnel mode SA is not modeled as an interface (e.g., as of this writing, popular in many open source implementations), the SPD entries for protecting all traffic between the two endpoints must be described. Evaluating against the requirements above, all link-local traffic multicast traffic would need to be identified, possibly resulting in a long list of SPD entries. The second requirement is difficult to satisfy, because the traffic needing protection is not necessarily (e.g., router-to-router tunnel) known a priori . The third requirement is also problematic, because almost all implementations assume addresses are assigned on interfaces (rather than configured in SPDs) for proper source address selection. If the IPsec tunnel mode SA is modeled as interface, the traffic that needs protection can be modeled as routes pointing to the interface. But the second requirement is difficult to satisfy, because the traffic needing protection is not necessarily known a priori. The third requirement is easily solved, because IPsec is modeled as an interface. In practice, (2) has been solved by protecting all the traffic (::/0), but no interoperable implementations support this feature. For a detailed list of issues pertaining to this, see . Because applying transport mode to protect a tunnel is a much simpler solution and also easily protects link-local and multicast traffic, we do not recommend using tunnel mode in this context. Tunnel mode is, however, discussed further in . This document assumes that tunnels are manually configured on both sides and the ingress filtering is manually set up to discard spoofed packets.

Transport mode has typically been applied to L2TP, GRE, and other tunneling methods, especially when the user wants to tunnel non-IP traffic. , , and provide examples of applying transport mode to protect tunnel traffic that spans only a part of an end-to-end path. IPv6 ingress filtering must be applied on the tunnel interface on all the packets that pass the inbound IPsec processing. The following SPD entries assume that there are two routers, Router1 and Router2, with tunnel endpoint IPv4 addresses denoted IPV4-TEP1 and IPV4-TEP2, respectively. (In other scenarios, the SPDs are set up similarly.)

The packet format is as shown in . Components (first to last) Contains IPv4 header (src = IPV4-TEP1, dst = IPV4-TEP2) ESP header IPv6 header (src = IPV6-EP1, dst = IPV6-EP2) (payload) The IDci and IDcr payloads of IKEv1 carry the IPv4-TEP1, IPV4-TEP2, and protocol value 41 as phase 2 identities. With IKEv2, the traffic selectors are used to carry the same information.

The Peer Authorization Database (PAD) provides the link between SPD and the key management daemon . This is defined in and hence relevant only when used with IKEv2. As there is currently no defined way to discover the PAD-related parameters dynamically, it is assumed that these are manually configured: The Identity of the peer asserted in the IKEv2 exchange: Many different types of identities can be used. At least, the IPv4 address of the peer should be supported. IKEv2 can authenticate the peer by several methods. Pre-shared key and X.509 certificate-based authentication is required by . At least, pre-shared key should be supported, because it interoperates with a larger number of implementations. The child SA authorization data should contain the IPv4 address of the peer. IPv4 address should be supported as Identity during the key exchange. As this does not provide Identity protection, main mode or aggressive mode can be used with IKEv1.

In , we examined the differences between setting up an IPsec IPv6-in-IPv4 tunnel using either transport or tunnel mode. We observe that applying transport mode to a tunnel interface is the simplest and therefore recommended solution. In , we also explore what it would take to use so-called Specific SPD (SSPD) tunnel mode. Such usage is more complicated because IPv6 prefixes need to be known a priori, and multicast and link-local traffic do not work over such a tunnel. Fragment handling in tunnel mode is also more difficult. However, because the Mobility and Multihoming Protocol (MOBIKE) supports only tunnel mode, when the IPv4 endpoints of a tunnel are dynamic and the other constraints are not applicable, using tunnel mode may be an acceptable solution. Therefore, our primary recommendation is to use transport mode applied to a tunnel interface. Source address spoofing can be limited by enabling ingress filtering on the tunnel interface. Manual keying must not be used as large amounts of IPv6 traffic may be carried over the tunnels and doing so would make it easier for an attacker to recover the keys. IKEv1 or IKEv2 must be used for establishing the IPsec SAs. IKEv2 should be used where supported and available; if not, IKEv1 may be used instead.

When running IPv6-in-IPv4 tunnels (unsecured) over the Internet, it is possible to "inject" packets into the tunnel by spoofing the source address (data plane security), or if the tunnel is signaled somehow (e.g., using authentication protocol and obtaining a static v6 prefix), someone might be able to spoof the signaling (control plane security). The IPsec framework plays an important role in adding security to both the protocol for tunnel setup and data traffic. Either IKEv1 or IKEv2 provides a secure signaling protocol for establishing, maintaining, and deleting an IPsec tunnel. IPsec, with ESP, offers integrity and data origin authentication, confidentiality, and optional (at the discretion of the receiver) anti-replay features. Using confidentiality without integrity is discouraged. ESP furthermore provides limited traffic flow confidentiality. IPsec provides access control mechanisms through the distribution of keys and also through the application of policies dictated by the Security Policy Database (SPD). The NAT traversal mechanism provided by IKEv2 introduces some weaknesses into IKE and IPsec. These issues are discussed in more detail in . Please note that using IPsec for the scenarios described in Figures , , and does not aim to protect the end-to-end communication. It protects just the tunnel part. It is still possible for an IPv6 endpoint not attached to the IPsec tunnel to spoof packets.

The authors are listed in alphabetical order. Suresh Satapati also participated in the initial discussions on this topic.

The authors would like to thank Stephen Kent, Michael Richardson, Florian Weimer, Elwyn Davies, Eric Vyncke, Merike Kaeo, Alfred Hoenes, Francis Dupont, and David Black for their substantive feedback. We would like to thank Pasi Eronen for his text contributions and suggestions for improvement.

&RFC.4213; &RFC.4306; &RFC.4301; &RFC.3948; &RFC.4303; BBN Corporation

70 Fawcett Street Cambridge MA 02140 USA +1 (617) 873-3988 kent@bbn.com

@Home Network

425 Broadway Redwood City CA 94063 USA +1 (415) 569-5000 rja@corp.home.net

Security IP security protocol IPSEC internet protocol version 6 security &RFC.2893; &RFC.3056; To be able to automate setting up IPv6-in-IPv4 tunnels, it is important to be able to automatically determine the tunnel end-point for the tunneling mechanism. This memo presents a short analysis and conclusions on the different approaches for discovering the IPv6 tunnel endpoint on a node. &RFC.4555; &RFC.4302; &RFC.4023; &RFC.3193; &RFC.3884; &RFC.4718;

First, we describe the different tunnel mode implementation methods. We note that, in this context, only the so-called Specific SPD (SSPD) model (without a tunnel interface) can be made to work, but it has reduced applicability, and the use of a transport mode tunnel is recommended instead. However, we will describe how the SSPD tunnel mode might look if one would like to use it in any case.

Tunnel mode could (in theory) be deployed in two very different ways depending on the implementation: "Generic SPDs": some implementations model the tunnel mode SA as an IP interface. In this case, an IPsec tunnel interface is created and used with "any" addresses ("::/0 <-> ::/0" ) as IPsec traffic selectors while setting up the SA. Though this allows all traffic between the two nodes to be protected by IPsec, the routing table would decide what traffic gets sent over the tunnel. Ingress filtering must be separately applied on the tunnel interface as the IPsec policy checks do not check the IPv6 addresses at all. Routing protocols, multicast, etc. will work through this tunnel. This mode is similar to transport mode. The SPDs must be interface-specific. However, because IKE uses IPv4 but the tunnel is IPv6, there is no standard solution to map the IPv4 interface to IPv6 interface and this approach is not feasible. "Specific SPDs": some implementations do not model the tunnel mode SA as an IP interface. Traffic selection is based on specific SPD entries, e.g., "2001:db8:1::/48 <-> 2001:db8:2::/48". As the IPsec session between two endpoints does not have an interface (though an implementation may have a common pseudo-interface for all IPsec traffic), there is no Duplicate Address Detection (DAD), Multicast Listener Discovery (MLD), or link-local traffic to protect; multicast is not possible over such a tunnel. Ingress filtering is performed automatically by the IPsec traffic selectors. Ingress filtering is guaranteed by IPsec processing when option (2) is chosen, whereas the operator has to enable it explicitly when transport mode or option (1) is chosen. In summary, there does not appear to be a standard solution in this context for the first implementation approach. The second approach can be made to work, but is only applicable in host-to-host or site-to-router/router-to-site scenarios (i.e., when the IPv6 prefixes can be known a priori), and it offers only a limited set of features (e.g., no multicast) compared with a transport mode tunnel. When tunnel mode is used, fragment handling may also be more difficult compared with transport mode and, depending on implementation, may need to be reflected in SPDs.

The following SPD entries assume that there are two hosts, Host1 and Host2, whose IPv6 addresses are denoted IPV6-EP1 and IPV6-EP2 (global addresses), and the IPV4 addresses of the tunnel endpoints are denoted IPV4-TEP1 and IPV4-TEP2, respectively.

The IDci and IDcr payloads of IKEv1 carry the IPV6-EP1 and IPV6-TEP2 as phase 2 identities. With IKEv2, the traffic selectors are used to carry the same information.

The following SPD entries assume that the host has the IPv6 address IPV6-EP1 and the tunnel endpoints of the host and router are IPV4-TEP1 and IPV4-TEP2, respectively. If the tunnel is between a router and a host where the router has allocated an IPV6-PREF/48 to the host, the corresponding SPD entries can be derived by replacing IPV6-EP1 with IPV6-PREF/48. Please note the bypass entry for host's SPD, absent in router's SPD. While this might be an implementation matter for host-to-router tunneling, having a similar entry, "Local=IPV6-PREF/48 & Remote=IPV6-PREF/48", is critical for site-to-router tunneling.

The IDci and IDcr payloads of IKEv1 carry the IPV6-EP1 and ID_IPV6_ADDR_RANGE or ID_IPV6_ADDR_SUBNET as their phase 2 identities. The starting address is zero and the end address is all ones for ID_IPV6_ADDR_RANGE. The starting address is zero IP address and the end address is all zeroes for ID_IPV6_ADDR_SUBNET. With IKEv2, the traffic selectors are used to carry the same information.

With the exchange of protected configuration payloads, IKEv2 is able to provide the IKEv2 peer with Dynamic Host Configuration Protocol (DHCP)-like information payloads. These configuration payloads are exchanged between the IKEv2 initiator and responder. This could be used (for example) by the host in the host-to-router scenario to obtain an IPv6 address from the ISP as part of setting up the IPsec tunnel mode SA. The details of these procedures are out of scope for this memo.

Network address (and port) translation devices are commonly found in today's networks. A detailed description of the problem and requirements of IPsec-protected data traffic traversing a NAT is provided in . IKEv2 can detect the presence of a NAT automatically by sending NAT_DETECTION_SOURCE_IP and NAT_DETECTION_DESTINATION_IP payloads in the initial IKE_SA_INIT exchange. Once a NAT is detected and both endpoints support IPsec NAT traversal extensions, UDP encapsulation can be enabled. More details about UDP encapsulation of IPsec-protected IP packets can be found in . For IPv6-in-IPv4 tunneling, NAT traversal is interesting for two reasons: One of the tunnel endpoints is often behind a NAT, and configured tunneling, using protocol 41, is not guaranteed to traverse the NAT. Hence, using IPsec tunnels would enable one to set up both a secure tunnel and a tunnel that might not always be possible without other tunneling mechanisms. Using NAT traversal allows the outer address to change without having to renegotiate the SAs. This could be beneficial for a crude form of mobility and in scenarios where the NAT changes the IP addresses frequently. However, as the outer address may change, this might introduce new security issues, and using tunnel mode would be most appropriate. When NAT is not applied, the second benefit would still be desirable. In particular, using manually configured tunneling is an operational challenge with dynamic IP addresses, because both ends need to be reconfigured if an address changes. Therefore, an easy and efficient way to re-establish the IPsec tunnel if the IP address changes would be desirable. MOBIKE provides a solution when IKEv2 is used, but it only supports tunnel mode.

The IKEv2 initiator needs to know the address of the IKEv2 responder to start IKEv2 signaling. A number of ways can be used to provide the initiator with this information, for example: Using out-of-band mechanisms, e.g., from the ISP's Web page. Using DNS to look up a service name by appending it to the DNS search path provided by DHCPv4 (e.g., "tunnel-service.example.com"). Using a DHCP option. Using a pre-configured or pre-determined IPv4 anycast address. Using other, unspecified or proprietary methods. For the purpose of this document, it is assumed that this address can be obtained somehow. Once the address has been learned, it is configured as the tunnel endpoint for the configured IPv6-in-IPv4 tunnel. This problem is also discussed at more length in . However, simply discovering the tunnel endpoint is not sufficient for establishing an IKE session with the peer. The PAD information (see ) also needs to be learned dynamically. Hence, currently, automatic endpoint discovery provides benefit only if PAD information is chosen in such a manner that it is not IP-address specific.