Source Address Validation in Intra-domain Networks Gap Analysis, Problem Statement, and Requirements

Introduction Source Address Validation (SAV) is important for defending against source address spoofing attacks. A multi-fence architecture called Source Address Validation Architecture (SAVA) was proposed to implement SAV at three levels: access-network SAV, intra-domain SAV, and inter-domain SAV (the "domain" used and discussed in this document means the Autonomous System (AS)). The multi-fence architecture helps enhance the effectiveness of SAV across the whole Internet by preventing or mitigating source address spoofing at multiple levels. Access-network SAV mechanisms (such as SAVI , IP Source Guard (IPSG) , and Cable Source-Verify ) can ensure that a host must use the source IP address assigned to the host. Given numerous access networks managed by different operators throughout the world, it is difficult to require all access networks to deploy SAV. Therefore, intra-domain SAV and inter-domain SAV are needed to block source-spoofed data packets from access networks as close to the source as possible. Intra-domain SAV and inter-domain SAV perform SAV at the granularity of IP prefixes, which is less granular than the access network SAV (i.e., IP address), as an IP prefix covers a range of IP addresses. This document focuses on the analysis of intra-domain SAV. In contrast to inter-domain SAV, intra-domain SAV does not require collaboration between different ASes. Intra-domain SAV rules can be generated by the AS itself. Intra-domain SAV for an AS aims at achieving two goals: i) blocking source-spoofed packets originated from its host networks or customer networks using a source address of other networks; and ii) blocking source-spoofed packets coming from other ASes using a source address of this AS. illustrates the goals and function of intra-domain SAV with two cases. Case i shows that the host network or customer network of AS X originates source-spoofed packets using a source address of other networks. If AS X deploys intra-domain SAV, the spoofed packets can be blocked by host-facing routers or customer-facing routers of AS X (i.e., Goal i). Case ii shows that AS X receives source-spoofed packets using a source address of AS X from other ASes (e.g., AS Y). If AS X deploys intra-domain SAV, the spoofed packets from AS Y can be blocked by AS border routers of AS X (i.e., Goal ii).

An example for illustrating intra-domain SAV | AS X | +-------------------------------+ +------+ Case ii: AS X receives packets spoofing source addresses of AS X from other ASes (e.g., AS Y) Goal ii: If AS X deploys intra-domain SAV, the spoofed packets can be blocked by AS X Spoofed packets using source addresses +------+ of AS X +------+ | AS X |<----------------------| AS Y | +------+ +------+ ]]> The scope of intra-domain SAV includes all IP-encapsulated scenarios:

Native IP forwarding: including both forwarding based on global routing table and VPN forwarding.
IP-encapsulated Tunnel (IPsec , GRE , SRv6 , etc.): focusing on the validation of the outer layer IP address.
Validating both IPv4 and IPv6 addresses.

Scope does not include:

Non-IP packets: including MPLS label-based forwarding and other non-IP-based forwarding.

In the following, this document provides gap analysis of existing intra-domain SAV mechanisms, concludes main problems of existing intra-domain SAV mechanisms, and proposes requirements for future ones.

Terminology SAV Rule: The rule in a router that describes the mapping relationship between a source address (prefix) and the valid incoming interface(s). It is used by a router to make SAV decisions. Host-facing Router: An intra-domain router facing an intra-domain host network. Customer-facing Router: An intra-domain router facing an intra-domain customer network. AS Border Router: An intra-domain router facing an external AS. Improper Block: The validation results that the packets with legitimate source addresses are blocked improperly due to inaccurate SAV rules. Improper Permit: The validation results that the packets with spoofed source addresses are permitted improperly due to inaccurate SAV rules. SAV-specific Information: The information specialized for SAV rule generation, which is exchanged among routers.

Requirements Language The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Existing Intra-domain SAV Mechanisms This section introduces existing intra-domain SAV mechanisms, including BCP38 and BCP84 .

ACL-based ingress filtering or BCP38 requires that network operators manually configure ACL rules on intra-domain routers to block or permit data packets using specific source addresses. This mechanism can be used on interfaces of host-facing or customer-facing routers facing an intra-domain host/customer network to prevent the corresponding host/customer network from spoofing source prefixes of other networks . In addition, it is also usually used on interfaces of AS border routers facing an external AS to block data packets using disallowed source addresses, such as internal source addresses owned by the local AS . In any application scenario, ACL rules must be updated in time to be consistent with the latest filtering criteria when the network changes.
Strict uRPF is also typically used on interfaces of host-facing or customer-facing routers facing an intra-domain host/customer network. Routers deploying strict uRPF accept a data packet only when i) the local FIB contains a prefix covering the packet's source address and ii) the corresponding outgoing interface for the prefix in the FIB matches the packet's incoming interface. Otherwise, the packet will be blocked.
Loose uRPF uses a looser validation method. A packet will be accepted if the router's local FIB contains a prefix covering the packet's source address regardless of the interface from which the packet is received. In fact, interfaces of AS border routers facing an external AS may use loose uRPF to block incoming data packets using non-global addresses .
Carrier Grade NAT has some operations on the source addresses of packets, but it is not an anti-spoofing tool, as described in . If the source address of a packet is in the INSIDE access list, the NAT rule can translate the source address to an address in the pool OUTSIDE. The NAT rule cannot determine whether the source address is spoofed or not. In addition, the packet using a spoofed source address will still be forwarded if the spoofed source address is not included in the INSIDE access list. Therefore, Carrier Grade NAT cannot help identify and block source-spoofed data packets.

Gap Analysis This section elaborates the key problems of current intra-domain SAV on host-facing or customer-facing routers and SAV on AS border routers, respectively.

SAV on Host-facing or Customer-facing Routers Towards Goal i in , intra-domain SAV is typically deployed on interfaces of host-facing or customer-facing routers facing an intra-domain host/customer network to validate data packets originated from that network, since SAV is more effective when deployed closer to the source. ACL-based ingress filtering and strict uRPF are commonly used for this purpose. ACL rules must be manually updated according to prefix changes or topology changes in a timely manner. Otherwise, if ACL rules are not updated in time, improper block or improper permit problems may occur. To ensure the accuracy of ACL rules in dynamic networks, high operational overhead will be induced to achieve timely updates for ACL configurations. Strict uRPF can generate and update SAV rules in an automatic way but it will cause improper blocks in the scenario of asymmetric routing or hidden prefix.

Asymmetric Routing shows asymmetric routing in a multi-homing scenario. In the figure, Network 1 is a host/customer network of the AS. It owns prefix 2001:db8::/32 and is attached to two intra-domain routers, i.e., Router 1 and Router 2. For the load balance purpose of traffic flowing to Network 1, Network 1 expects the incoming traffic destined for the sub-prefix 2001:db8:8000::/33 to come only from Router 1 and the incoming traffic destined for the other sub-prefix 2001:db8::/33 to come only from Router 2. To this end, Router 1 only learns the route to sub-prefix 2001:db8:8000::/33 from Network 1, while Router 2 only learns the route to the other sub-prefix 2001:db8::/33 from Network 1. Then, Router 1 and Router 2 distribute the sub-prefix information to routers in the AS through intra-domain routing protocols such as OSPF or IS-IS. The FIBs of Router 1 and Router 2 are shown in the figure. Although Network 1 does not expect traffic destined for 2001:db8::/33 to come from Router 1, it may send traffic with source addresses of prefix 2001:db8::/33 to Router 1 for load balance of traffic originated from Network 1. As a result, there is asymmetric routing of data packets between Network 1 and Router 1. Arrows in the figure indicate the flowing direction of traffic. In addition to the traffic engineering mentioned above, other factors may also cause the similar asymmetric routing between host-facing/customer-facing routers and host/customer networks.

Asymmetric routing in multi-homing scenario If Router 1 uses strict uRPF on interface '#', the SAV rule is that Router 1 only accepts packets with source addresses of 2001:db8:8000::/33 from Network 1. Therefore, when Network 1 sends packets with source addresses of 2001:db8::/33 to Router 1, strict uRPF at Router 1 will improperly block these legitimate packets. Similarly, when Router 2 uses strict uRPF on its interface '#' and receives packets with source addresses of prefix 2001:db8:8000::/33 from Network 1, it will also improperly block these legitimate packets because strict uRPF at Router 2 will only accept packets from Network 1 using source addresses of prefix 2001:db8::/33.

Hidden Prefix For special business purposes, a host/customer network will originate data packets using a source address that is not distributed through intra-domain routing protocol. In other words, the IP address/prefix is hidden from intra-domain routing protocol and intra-domain routers. In this scenario, strict uRPF on host-facing or customer-facing routers will improperly block data packets from the host/customer network using source addresses in a hidden prefix.

Hidden prefix in CDN and DSR scenario The Content Delivery Networks (CDN) and Direct Server Return (DSR) scenario is a representative example of hidden prefix. In this scenario, the edge server in an AS will send DSR response packets using a source address of the anycast server which is located in another remote AS. However, the source address of anycast server is hidden from intra-domain routing protocol and intra-domain routers in the local AS. While this is an inter-domain scenario, we note that DSR response packets may also be improperly blocked by strict uRPF when edge server is located in the host/customer network. For example, in , assume edge server is located in Host/Customer Network 2 and Router 5 only learns the route to P2 from Network 2. When edge server receives the request from the remote anycast server, it will directly return DSR response packets using the source address of anycast server (i.e., P3). If Router 5 uses strict uRPF on interface '#', the SAV rule is that Router 5 only accepts packets with source addresses of P2 from Network 2. As a result, DSR response packets will be blocked by strict uRPF on interface '#'. In addition, loose uRPF on this interface will also improperly block DSR response packets if prefix P3 is not in the FIB of Router 5.

SAV on AS Border Routers Towards Goal ii in , intra-domain SAV is typically deployed on interfaces of AS border routers facing an external AS to validate packets arriving from other ASes. shows an example of SAV on AS border routers. In the figure, Router 3 and Router 4 deploy intra-domain SAV on interface '#' for validating data packets coming from external ASes. ACL-based ingress filtering and loose uRPF are commonly used for this purpose.

An example of SAV on AS border routers By configuring ACL rules, data packets that use disallowed source addresses (e.g., non-global addresses or internal source addresses) can be blocked at AS border routers. However, the operational overhead of maintaining updated ACL rules will be extremely high when there are multiple AS border routers adopting SAV as shown in . As for loose uRPF, it sacrifices the directionality of SAV and has limited blocking capability, because it allows packets with source addresses that exist in the FIB table on all router interfaces.

Problem Statement As analyzed above, existing intra-domain SAV mechanisms have significant limitations on automatic update or accurate validation. ACL-based ingress filtering relies on manual configurations and thus requires high operational overhead in dynamic networks. To guarantee accuracy of ACL-based SAV, network operators have to manually update the ACL-based filtering rules in time when the prefix or topology changes. Otherwise, improper block or improper permit problems may appear. Strict uRPF can automatically update SAV rules, but may improperly block legitimate traffic under asymmetric routing scenario or hidden prefix scenario. It may mistakenly consider a valid incoming interface as invalid, resulting in improper block problems; or it may mistakenly consider an invalid incoming interface as valid, resulting in improper permit problems. Loose uRPF is also an automated SAV mechanism but its SAV rules are overly loose. Most spoofed packets will be improperly permitted by loose uRPF. In summary, uRPF cannot guarantee the accuracy of SAV because it solely uses the router’s local FIB or RIB information to determine SAV rules. A router cannot see the asymmetric route between itself and another router/network from its own perspective. As a result, strict uRPF has improper block problems in the scenario of asymmetric route. The network operator has a comprehensive perspective so he/she can configure the correct SAV rules. However, manual configuration has limitations on automatic update.

Requirements for New SAV Mechanisms This section lists the following five requirements and related discussions that should be considered when designing the new intra-domain SAV mechanism.

Accurate Validation The new intra-domain SAV mechanism needs to improve the validation accuracy upon existing intra-domain SAV mechanisms. In a static network, improper block MUST be avoided to guarantee that legitimate traffic will not be blocked. Improper permit SHOULD be reduced as much as possible so that data packets with spoofed source addresses can be effectively blocked. When the network changes, the new mechanisms MUST efficiently update SAV rules to guarantee the accuracy of SAV. The key to overcoming the improper block problems of uRPF is to combine perspectives of multiple routers, allowing each router to form a more comprehensive perspective. To this end, the new SAV mechanism can allow routers to exchange and update the necessary information (i.e., SAV-specific information) that will affect the accuracy of SAV. By integrating SAV-specific information of multiple routers, routers can determine accurate SAV rules in the scenario of forwarding or routing asymmetry. To improve compatibility with the current intra-domain network, it is recommended that existing routing protocols can be used to implement the SAV function.

Automatic Update The new intra-domain SAV mechanism MUST be able to automatically generate and update SAV rules based on SAV-specific information or/and routing information received from other routers, instead of purely relying on manual update. While manual configuration may still be required in some special cases (such as in the hidden prefix scenario where only the operator knows the hidden prefix), automation can reduce considerable operational overhead in general.

Working in Incremental Deployment The new intra-domain SAV mechanism SHOULD NOT assume pervasive adoption. Since routers scheduled to adopt the new mechanism may not be able to be upgraded simultaneously, the new intra-domain SAV mechanism SHOULD be able to provide incremental protection under incremental deployment. Specifically, it SHOULD specify the scope of deployment and provide clear incremental benefits so that network operators can decide which routers to deploy SAV on first by evaluating the effectiveness of blocking spoofing traffic. In addition, it SHOULD outperform the current ones in the same incremental deployment scenario.

Fast Convergence The new intra-domain SAV mechanism MUST adapt to prefix changes, route changes, and topology changes in an intra-domain network, and update SAV rules in a timely manner. It MUST consider how to update SAV rules proactively or reactively so as to minimize improper blocks during convergence. It is also important to reduce the complexity and overhead of SAV implementation. To this end, routers MUST NOT signal too much information. Only information that cannot be learned from local routing information and is necessary for SAV should be signaled.

Necessary Security Guarantee Necessary security tools SHOULD be considered in the new intra-domain SAV mechanism. These security tools can help protect the SAV rule generation process. details the security scope and considerations for the new intra-domain SAV mechanism.

Security Considerations The new intra-domain SAV mechanisms should not introduce additional security vulnerabilities or confusion to the existing intra-domain architectures or control or management plane protocols. Similar to the security scope of intra-domain routing protocols, intra-domain SAV mechanisms should ensure integrity and authentication of protocol messages that deliver the required SAV information, and consider avoiding unintentional misconfiguration. It is not necessary to provide protection against compromised or malicious intra-domain routers which poison existing control or management plane protocols. Compromised or malicious intra-domain routers may not only affect SAV, but also disrupt the whole intra-domain routing domain. Security solutions to prevent these attacks are beyond the capability of intra-domain SAV.

IANA Considerations This document does not request any IANA allocations.

Acknowledgements Many thanks to the valuable comments from: Jared Mauch, Barry Greene, Fang Gao, Kotikalapudi Sriram, Anthony Somerset, Yuanyuan Zhang, Igor Lubashev, Alvaro Retana, Joel Halpern, Aijun Wang, Michael Richardson, Li Chen, Gert Doering, Mingxing Liu, Libin Liu, John O'Brien, Roland Dobbins, Xiangqing Chang, Tony Przygienda, Yingzhen Qu, Changwang Lin, James Guichard, Linda Dunbar, Robert Sparks, Yu Fu, Stephen Farrel etc.