HomeIncident library
Incident library

INCIDENT REPLAY · 2024-06-27

Cloudflare 1.1.1.1 hijack and route leak (Eletronet, Nova Rede)

AS267613 originated 1.1.1.1/32 and, separately, AS262504 leaked 1.1.1.0/24 upstream where AS1031 propagated it. At least one Tier 1 treated the /32 as a blackhole route. Over 300 networks in 70 countries had 1.1.1.1 unreachable or very slow for about 7.5 hours.

Mis-originationBGP withdrawal / announcementCatalog confidence: high

A reconstruction, not the operator's network. Hammerhead's authors rebuilt the change from the public record as a small network of their own. No configuration here is the operator's. What follows is what the tool reported on that reconstruction; it is not a claim that the tool would have prevented the outage.

What happened

Who: Cloudflare (victim); AS267613 Eletronet, AS262504 Nova Rede, AS1031 Peer-1 Global IX

The summary at the top is paraphrased from the sources below. Hammerhead's catalog rates its confidence in this account high.

Sources

What the public record says

  • AS267613 originated 1.1.1.1/32 and, separately, AS262504 leaked 1.1.1.0/24 upstream where AS1031 propagated it. At least one Tier 1 treated the /32 as a blackhole route. Over 300 networks in 70 countries had 1.1.1.1 unreachable or very slow for about 7.5 hours.
  • Who: Cloudflare (victim); AS267613 Eletronet, AS262504 Nova Rede, AS1031 Peer-1 Global IX

What the reconstruction assumed

Decided by Hammerhead's authors to build a network that runs. None of it describes the operator's network.

Everything below is ours.

  • The family template (tools/replays/gen_family_misorigination.py): ELETRONET-1 (AS 267613), an upstream TIER1-1 (AS 64500) that prefers its customer's routes and filters nothing on that session, the victim, and a third party behind the upstream. ELETRONET-1 null-routes the space it originates, so traffic drawn to it is dropped.
  • Cisco IOS syntax. No operator here has published configuration, and which vendors they ran is not on the record.
  • The upstream's other sessions are filtered and limited, so that the audit's finding is about this session and not the template.
  • ELETRONET-1 exports with the ordinary customer policy, an AS-path filter that passes only what it originates (^$). A mis-originated prefix is locally originated, so it passes.
  • The Tier 1 is not named by Cloudflare. It accepted the /32 under a remote-triggered-blackhole policy; here it simply accepts what its customer sends. The separate leak of 1.1.1.0/24 by Nova Rede is not modelled.

What Hammerhead said

Each run below is pinned by Hammerhead's replay tests at commit 0ee9732: the verdict and the quoted text are checked against the tool's output, and a run that disagrees fails the test. Reason lines are copied verbatim; a pinned line can be part of a longer reason. Every preflight run uses --max-blast-radius 100, so a change is never counted as caught merely because it touched many devices of a small network.

every network in one snapshot

now: fail

On the reconstruction, hammerhead preflight on a before and an after snapshot returned fail, for the reasons below.

Reason text pinned (some reason contains each line)

  • ELETRONET-1 would start originating 1.1.1.1/32, which is inside 1.1.1.0/24 that it learns from AS 64500, originated by AS 13335
  • connected subnet(s) reachable before are not reachable after

First run, before the fixes this batch prompted

first: fail
  • reachability regression: 1 connected subnet(s) reachable before are not reachable after — 1.1.1.1/32 on CLOUDFLARE-1 (1.1.1.1) from ELETRONET-1, TIER1-1, X-1 [icmp echo, tcp/179 bgp, tcp/22 ssh, tcp/80 http, udp/53 dns]
  • external advertisements: ELETRONET-1 would start originating 1.1.1.1/32, which is inside 1.1.1.0/24 that it learns from AS 64500, originated by AS 13335. A more-specific of address space you receive from a peer is not yours to announce

Pinned in tests/gauntlet/replays-misorigination.yaml.

ELETRONET-1's own router, nothing declared: a new origination is reported, and only a caveat

now: pass

On the reconstruction, hammerhead preflight on a before and an after snapshot returned pass.

Caveats pinned (some caveat contains each line; a caveat never moves the verdict)

  • ELETRONET-1 would start announcing 1 prefix(es) to

First run, before the fixes this batch prompted

first: unknown

Pinned in tests/gauntlet/replays-misorigination.yaml.

ELETRONET-1's own router, with what the upstream sends declared

now: fail

On the reconstruction, hammerhead preflight on a before and an after snapshot with --boundary @eletronet_2024/boundary.yaml returned fail, for the reasons below.

Reason text pinned (some reason contains each line)

  • ELETRONET-1 would start announcing 1.1.1.1/32 to
  • which is inside 1.1.1.0/24 that
  • (AS 64500) advertises to this network (declared by

First run, before the fixes this batch prompted

first: fail
  • external advertisements: ELETRONET-1 would start announcing 1.1.1.1/32 to 172.16.1.1 (AS 64500), which is inside 1.1.1.0/24 that 172.16.1.1 (AS 64500) advertises to this network (declared by noc@example.net). A more-specific of address space you receive from a peer is not yours to announce

Pinned in tests/gauntlet/replays-misorigination.yaml.

ELETRONET-1's own router, with the validator's ROAs

now: fail

On the reconstruction, hammerhead preflight on a before and an after snapshot with --roas @eletronet_2024/roas.json returned fail, for the reasons below.

Reason text pinned (some reason contains each line)

  • RPKI: ELETRONET-1 would start originating 1.1.1.1/32 as AS 267613
  • authorises AS 13335

First run, before the fixes this batch prompted

first: not runnable

error: unexpected argument '--roas' found

Pinned in tests/gauntlet/replays-misorigination.yaml.

TIER1-1's own router, no change: the audit

now: fail (finding)

An audit of one snapshot, no change: hammerhead nqe-library --mode run, check nqe.bgp.ebgp_peer_unfiltered_and_unlimited. Pinned status: fail, a finding whose evidence contains the text below.

Evidence pinned (the evidence contains each line)

  • TIER1-1
  • 267613

First run, before the fixes this batch prompted

first: pass (no finding)

Pinned in tests/gauntlet/replays-misorigination.yaml.

every network in one snapshot (potential microloop)

now: fail

On the reconstruction, hammerhead preflight on a before and an after snapshot returned fail.

This run repeats another pinned run (same snapshots, flags and verdict) to record the potential microloops it finds. A caveat never moves the verdict.

Caveats pinned (some caveat contains each line; a caveat never moves the verdict)

  • ELETRONET-1 -> TIER1-1 -> ELETRONET-1, if TIER1-1 has its new next hop while ELETRONET-1 still has its old one

Pinned in tests/gauntlet/replays-transients.yaml.

What changed in Hammerhead because of it

From the “What was fixed” table in Hammerhead's replay notes, where each fix landed after a failing test. “Found by” is copied as written; a family name means the batch this incident belongs to.

ROA maximum length

Roa::covers_prefix refused to cover a more-specific beyond maxLength, so such a prefix came back NotFound instead of Invalid (RFC 6811).

Found by: Mis-origination family

--roas <FILE> origin-validates new originations

From the originator's own seat a block that escapes is one new prefix, and only a caveat. With a validator's export it is decidable: "the ROA for 104.244.42.0/24 authorises AS 13414".

Found by: Mis-origination family

An announcement that multiplies

The announcement to an external peer goes from a few prefixes to many in one change.

Found by: Mis-origination family, Level 3

Outbound policy runs before the local AS is prepended

The simulator prepended first, so an outbound filter-list with ^$ matched nothing the router originated.

Found by: Mis-origination family

Many foreign originations are one reason

A mass mis-origination of hundreds of prefixes was hundreds of reasons.

Found by: Mis-origination family

Limits

  • The configurations are a reconstruction. Where the record is silent, the assumptions above fill the gap; if one is wrong, the replay is wrong in that respect.
  • A verdict describes the reconstructed network's converged state. Timing, traffic load, and anything a device does that no configuration states, such as software reacting to a health check or a firmware bug, are outside it.
  • A pass is reported as a pass. A fail says a gate fired on the reconstruction, nothing more.
  • Why the catalog counts this outage as in scope: Three static findings: a /32 of a placeholder address reaching eBGP export; a blackhole-community policy that accepts RTBH for space the customer does not own; and no /24 maximum length or ROV on the accepting sessions.

Not fixed: A single-prefix hijack from the hijacker's own seat, with no ROA and nothing declared

One new origination is what a legitimate new prefix looks like. pass with the caveat naming it; --roas or --boundary decides it.

Recorded against: Mis-origination family (17 of 28)

Not fixed: A forged-origin hijack

Origin validation passes it by construction. ASPA is not modelled.

Recorded against: Mis-origination family

Imported from docs/replays/eletronet-2024/FACTS.md and the replay manifests in the Hammerhead repository at commit 0ee9732.