Skip to content
AdminKit

BCP 38 (Source Address Validation)

What is BCP 38 (Source Address Validation)?

BCP 38, published as RFC 2827, is the IETF best-practice rule that network operators filter outbound traffic at the network edge, dropping any packet whose source address isn't actually authorized for that network. AdminKit's My IP tool shows your own visible address -- exactly what BCP 38 filtering at your edge stops someone else from forging.

By AdminKit team, Netbase JSC · Updated

BCP 38, published as RFC 2827 (tracked on the IETF Datatracker, with RFC 3704 extending it to multihomed networks), is the Internet Engineering Task Force’s best-practice recommendation that network operators filter outbound traffic at the network edge, dropping any packet whose source address doesn’t belong to a prefix that network is actually authorized to originate. Without this ingress filtering, a host on any network can send a packet claiming almost any source address it likes — a different kind of forgery than ARP spoofing’s local, link-layer trickery, or traffic simply appearing to come from a bogon address that was never allocated at all — which is exactly what spoofed-source denial-of-service and reflection/amplification attacks depend on: the attacker sends small requests with a victim’s address forged as the source, and every reply goes to the victim instead of the attacker — Memcached’s UDP interface is a frequently cited example, with an amplification factor CISA has measured at up to 51,000x. BCP 38 doesn’t stop an attacker from spoofing an address that is legitimately inside their own permitted range, and it only works when deployed broadly — a network that filters its own outbound spoofed traffic protects the rest of the internet from attacks it would have originated, not necessarily itself from attacks originating elsewhere. To make that deployment measurable, CAIDA’s Spoofer project maintains open-source client software for Windows, macOS and UNIX-like systems that periodically tests a network’s ability to send packets with forged source addresses, and reports on how widely source address validation is deployed. Adoption remains uneven decades after publication, since the benefit of filtering accrues to the rest of the internet rather than to the operator doing the work.

Why BCP 38 matters in practice:

  • Stops spoofed-source traffic at the edge, before it reaches the open internet.
  • Doesn’t protect the filtering network itself from attacks launched elsewhere.
  • Only works at internet scale when deployed broadly, which is why adoption still lags decades after RFC 2827.

BCP 38 (Source Address Validation) is part of AdminKit’s IP & network cluster; the full AdminKit glossary lists other related terms.

Frequently asked questions

What does BCP 38 require a network operator to do?

Filter outbound traffic at the network edge, dropping any packet whose source address doesn't belong to a prefix that network is actually authorized to originate.

What attack does BCP 38 filtering prevent?

Spoofed-source denial-of-service and reflection/amplification attacks, where an attacker sends small requests with a victim's address forged as the source so every reply goes to the victim instead.

Does deploying BCP 38 protect the network that deploys it?

Not directly -- a network that filters its own outbound spoofed traffic protects the rest of the internet from attacks it would have originated, not necessarily itself from attacks originating elsewhere.

What extends BCP 38 to multihomed networks?

RFC 3704, which addresses ingress filtering for networks with more than one upstream provider, where RFC 2827's original approach doesn't apply cleanly.