Source Address Validation Improvements
|Document||Charter||Source Address Validation Improvements WG (savi)|
|Title||Source Address Validation Improvements|
|IESG||Responsible AD||Suresh Krishnan|
|Charter edit AD||(None)|
|Send notices to||(None)|
While ingress filtering [RFC 2827, BCP 38] provides a way to validate IP source addresses at an aggregated level, there is not yet a standardized mechanism for IP source address validation at a finer granularity. Having a finer granularity would be helpful in a number of situations, including filtering traffic from customer interfaces implemented as ports in a layer 3 aware bridge or a router, general improvements in filtering accuracy in enterprise networks, etc. Depending on the situation, there may be a requirement for blocking spoofed packets or merely logging packets that appear to be spoofed. Partial solutions exist to prevent nodes from spoofing the IP source address of another node in the same IP link (e.g., the "IP source guard"), but are proprietary. The purpose of the proposed "Source Address Validation Improvements" working group is to standardize mechanisms that prevent nodes attached to the same IP link from spoofing each other's IP addresses. The scope of the WG is as follows: - The working group considers only solutions implemented on systems located on the same IP link as a to-be-verified node. The goal of the working group is the LAN environment and solutions running in routers or layer 3 aware Ethernet bridges. - Both IPv4 and IPv6 need to be covered. - The first goal of the working group is on unicast traffic, but using the same mechanisms to police multicast traffic is also within the scope. - All address assignment mechanisms need to be supported, including stateless, stateful, and manual configuration; as well as privacy and cryptographically generated addresses. - Solutions are preferably based on observing user traffic, or on observing or using existing signaling protocols. Examples of protocols that can be useful to observe/use are ARP, Neighbor Discovery, DHCP, and DHCP Prefix Delegation protocols. Observing addresses in IP headers can also be useful. The gathered information is used to determine what IP source addresses in packets are appropriate. Where automatic operation is impossible or would lead to sub-optimal validation results, solutions may require manual configuration. - Interdomain scenarios (across Autonomous Systems) that require information from routing protocols like BGP are out of scope. Nevertheless, solution may observe routing protocol signaling to detect that a device is a router. - Tracking other protocols is not within the scope of the WG. - No changes to hosts are allowed. - The WG is prohibited from creating its own protocols or extensions/modifications of current protocols. These limitations in the scope may be relaxed through later rechartering. For instance, solutions tailored for PPP links and specific environments may be added later, or solutions involving co-operation of the nodes on the link may be developed once the baseline solutions have been completed. However, the WG is already chartered to work also on a solution for Ethernet-based broadband access networks that are used in DSL environments. This work is a specialization of the working group's primary LAN-based solution. In order to reach a result that is widely usable and unlikely to disturb existing network practices, the working group needs to take into account - nodes that use static addresses, - nodes with multiple IP addresses on the same interface, - nodes that use multiple link-layer addresses on the same interface, - nodes that have multiple interfaces to the same link, - attachment of another bridge at a bridge port, - presence of routers, NATs, and other similar devices on the same link, including their distinction from hosts with multiple interfaces or hosts with multiple IP addresses on a single interface, - use of SEcure Neighbor Discovery in some networks, - nodes that move to another port on the same link, and - hosts with anycast addresses. However, should such wide applicability turn out to be impossible, the working group will document the limitations of the solutions in due manner. In particular, it is likely that anycast addressing and nodes that employ multiple interfaces for load balancing at link layer are indistinguishable from an actual spoofing attack. There may also be a difference in the applicability between blocking and merely logging spoofed packets. In any case, the solutions may require to be explicitly turned on for each network or interface where they are applicable. For background information, the working group will also develop a threats analysis document that describes what threats the solutions from the WG protect against. This document also contrasts SAVI to existing solutions.