IntraSAV - A Solution for Intra-Domain Source Address Validation

Introduction The scope of intra-domain SAV is stated in as follows: "Intra-domain SAV is applied at external interfaces (on routers) facing entities that are not deployed as neighboring ASes and are therefore not covered by inter-domain SAV. For example, an entity can be a single host, a set of hosts, or a customer network with no AS that manages one or more IP prefixes. The entity may source traffic using prefixes assigned by the AS or its own BYOIP prefixes. From the perspective of other ASes, such traffic is originated by the AS." This document specifies a solution, named IntraSAV, for intra-domain source address validation (SAV). This solution addresses the problem stated in the ietf-savnet-intra-domain-problem-statement (RFC-to-be) document.

Terminology The terminology used in this document follows Section 1.1 in .

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.

IntraSAV Solution The focus here is on SAV at router interfaces, where traffic sourced from prefixes used (or originated) at the local AS (i.e., the SAV-performing AS) arrives from either (1) directly connected hosts, or (2) customers with non-BGP interfaces . In these scenarios, the AS may support two types of prefixes for route or traffic origination: (1) its own provider-owned prefixes, and (2) Bring Your Own IP (BYOIP) prefixes. The proposed SAV solution requires BYOIP customers to inform the local AS about the specific prefixes they intend to use on their respective interfaces for routing and/or source addressing. This data becomes part of the configuration information at the local AS (as defined in the Terminology section of ). The local AS similarly incorporates the routing and source addressing requirements (on a per-interface basis) for its own (provider-owned) prefixes into this configuration. Finally, possibly using a SAV agent, the local AS uses the consolidated configuration information to construct and deploy an intra-domain SAV table (allowlist) for each relevant interface of its Customer Edge (CE) routers. The prefix owners (whether the BYOIP customers or the local AS) should register ROAs for their prefixes, authorizing the local AS as the origin. For prefixes originating at the local AS (for either routing or sourcing), the local configuration information takes precedence over the ROA information for both route origination and SAV. However, the local AS should still check the ROAs against its configuration and alert customers to any inconsistencies. This verification is important because remote ASes (one or more hops away) rely on these ROAs (along with other information) to compute their own inter-domain SAV tables . The IntraSAV solution method is illustrated below in . The figure also illustrates an internal AS topology. IntraSAV is applied at CE routers 1 and 2 on Interfaces 1 and 2 that support Customers 1 and 2, respectively. The Configuration Manager (CM), including a SAV Agent, plays a key role in facilitating IntraSAV. The role of the CM is to authenticate the customers and collect their configuration information for routing and SAV purposes. Customer 1 registers prefixes {p, q} and specifies that both prefixes are routed (i.e., announced). Note that "can be routed (announced)" implies that the prefix can also be used for sourcing traffic. On the other hand, Customer 2 registers prefixes {r, s} and specifies that r is routed while s is not routed but used for sourcing at this AS. The SAV agent in the CM makes use of this configuration information and programs CE router interfaces 1 and 2 with SAV tables (allowlists) {p, q} and {r, s}, respectively. As long as the configuration information available to the local Configuration Manager is complete, IntraSAV guarantees zero improper blocks and zero improper admits.

Illustration of internal AS topology and the IntraSAV solution method. ** To be Discussed (by authors and the WG): The AS operator may choose to apply IntraSAV on Interface 3 at the ASBR in . This type of ingress SAV may be done in addition to or as an alternative to doing IntraSAV at the CE router interfaces described above. The important consideration is that the operator should be certain, based on knowledge of the local topology and information in the CM, that all the traffic the ASBR interface receives is expected to be sourced from known prefixes and originating from the local AS. It is clear in the topology of that the expected set of traffic-sourcing prefixes at Interface 3 is equal to the superset of the sets of prefixes configured for the CE router interfaces (Interfaces 1 and 2). So, the operator can confidently apply the prefix-superset {p, q, r, s} as the SAV table (allowlist) on Interface 3 at the ASBR (see ). It may be noted that this proposal (i.e., SAV at an internal ASBR interface) likely goes beyond the scope stated in except when the ASBR interface in consideration is directly serving a set of hosts or a non-AS customer. Generally, this proposal is useful for cases when the AS operator determines that it is much more feasible to upgrade their ASBR rather than the CE routers for SAV. **

Implementation and Deployment Considerations The local AS maintains a Configuration Manager (CM). For the prefixes provisioned by the local AS, the AS operator enters in the CM the configuration information that associates the prefixes to specific interfaces for the purposes of route announcements and SAV. Non-AS customers with BYOIP prefixes also register configuration information for their prefixes in the CM. An example data structure for the configuration information is shown in . Each such record lists routing and sourcing-only prefixes applicable for each {customer, interface} pair. The interface ID is provided by the AS operator. The routed prefixes () are included by default in the SAV table for the given interface; the SAV agent () performs this work. If a set of sourcing-only prefixes is present in the record (), this set is also included in the SAV table. The configuration information in the CM is updated whenever any changes occur in prefixes regarding their routing or sourcing usage at an interface.

Illustration of a Configuration Manager data structure. A customer may use the bottommost field () to provide an optional special instruction such as "the union of all our prefixes must be included in the SAV table at all router interfaces where we are present (at this AS)". In this example, the customer's registered configuration information conveys that they may be routing different sets of prefixes from different interfaces but use a common union set of prefixes for source addresses at all those interfaces. This is a form of asymmetric routing and sourcing. The data structure illustrated in allows the customer to register (configure) any form of asymmetric routing or sourcing.

Security Considerations Checking the configuration information against ROA information helps in affirming accuracy and avoiding improper blocks and improper permits in IntraSAV at the local AS. This, in turn, helps with avoiding improper blocks and improper permits in inter-domain SAV (e.g., BAR-SAV ) at ASes one or more hops away. The AS operator or its CM should cross-check the customer ID (or organization name) and associated prefixes against the corresponding RPKI resource certificate or prefix registration in the RIR. This check serves as an enhanced form of authentication (reinforcing the customer registration process) for the configuration information for a given customer.

IANA Considerations This document does not have IANA considerations.

Acknowledgements The authors wish to thank Paul Vixie, Doug Montgomery, Jeff Haas, Joel Halpern, and Ron Bonica for suggestions and discussions.