AS401573 · Route server · Operational

Drop the threat before it reaches your firewall.

BlackHawk BlackHole is a BGP blackhole route server. Bring up one eBGP session and the routers you already own start discarding traffic to command-and-control servers, ransomware infrastructure, botnets and hostile networks — in hardware, at line rate, before a firewall rule or a proxy licence is ever consulted.

Then point the next-hop at a collector, and it tells you which of your machines is already infected.

Peering session — illustrative
neighbor198.51.100.10
remote-as401573
stateESTABLISHED
familiesIPv4 · IPv6
authTCP-MD5 (RFC 2385)
communities65535:666 · 401573:k:c
inbound policyreject all
BlackHawk Threat BGP Route Server. Live prefix, country and source counts are rendered from the routing table in the console — never baked into a page that would go stale.
12
Router platforms generated
19
Threat categories, individually toggled
~10s
Policy change to the wire, no session flap
1
eBGP session. No agent, tunnel or appliance
Why this is a different category

You are already paying to inspect traffic you could have thrown away.

Every control you own — next-generation firewall, IDS/IPS, secure web gateway, DNS filtering, EDR — inspects traffic in software, and every one of them is priced or performance-limited by how much traffic it inspects and how many indicators it holds. BlackHole does not inspect anything. It tells hardware you already own what to discard.

 Firewall / IPS / SWGBlackHawk BlackHole
Where the drop happensSoftware inspection pathForwarding plane (ASIC), at line rate
Marginal cost per dropped packetInspection budget, a state-table entry, a licence tierNone
Indicator list size limitVendor-imposed, often low thousandsA routing table — hundreds of thousands of prefixes is routine
DeploymentAn agent, an appliance, an inline tap, or a licenceOne eBGP session. No new hardware, no agent, no tunnel
Time to first valueA procurement cycleOne router config paste
Effect on inspection capacityConsumes itReturns it — the traffic never arrives to be inspected

Your routers can already discard millions of packets per second without breaking a sweat. BlackHole simply tells them what to discard.

This is not a firewall replacement. It is the capacity already sitting idle in your forwarding plane — and every packet it discards is one your firewall no longer has to inspect. We worked the arithmetic out here.

How it works

One session. Four things happen.

No appliance to rack, no agent to roll out, no tunnel to maintain. The whole deployment is a router configuration you paste.

01

One eBGP session

Multihop from a loopback on a router you already own, authenticated with a per-router TCP-MD5 key you generate and can rotate in the console. No tunnel, no agent, no appliance.

02

Threat prefixes arrive tagged

Every prefix carries the RFC 7999 blackhole community 65535:666 and an RFC 8092 large community identifying its category, so your own policy decides what to do with each class.

03

Your router discards in hardware

The import policy rewrites the next-hop to a discard route. From there the drop happens in the forwarding plane at line rate, before a firewall rule, an IDS signature or a proxy licence is consulted.

04

Or point it somewhere that watches

Aim the next-hop at an on-premises collector instead of null0 and each blocked beacon becomes a named internal machine and a named user, rather than an anonymous dropped packet.

What you gain

Nine things a routing table does that an indicator list cannot.

Line-rate, zero marginal cost

A forwarding-plane drop consumes no inspection budget, no state-table entry and no licence tier. It is the one place in the network where discarding a packet is genuinely free.

A routing table, not an indicator list

Inspection products cap indicators in the low thousands. Hundreds of thousands of prefixes is an ordinary day for a router.

IPv4 and IPv6 on one session

MP-BGP with correct next-hops for both families. IPv6 threat prefixes carry full parity, not an afterthought.

Policy changes without a flap

Toggling a category rewrites the export filter and issues an outbound route refresh. The session stays up and the routes change underneath it, within about ten seconds.

Egress-only, by architecture

One unconditional global policy rejects every route from every peer. We cannot accept a route from you, so your session cannot be used to leak or hijack a prefix.

Multi-node redundancy on every plan

Every router peers with every active peering node, on any plan. Retiring a node keeps existing sessions running and marks them draining. Redundancy is never the thing being sold.

Per-router keys, per-router overrides

Each router has its own address, ASN and MD5 key, deliberately not shared — one compromised box must not let an attacker impersonate the others.

Roles an auditor can live inside

Owner, admin, member and viewer, enforced in the database. Read-only access without the ability to blackhole a /8.

One alert per incident, not per edge

A session that drops and bounces thirty-nine times in an afternoon is one incident. Sending thirty-nine emails about it is how an operator mutes the alert and misses the next real outage.

Build your feed

Subscribe to exactly what you want dropped.

Every category is independently toggled and applies over the live session in about ten seconds, without a reconnect. Toggle them here and watch the configuration below change with you.

Start from
12of 19 categories selected
5 critical6 high1 operational

This selection is stamped into the configuration generated below, the way the console stamps your real one. Prefix counts per category are deliberately not shown here — they change continuously and are rendered live in the console, where they will always agree with what your session is actually carrying.

Per-country blocking, and the honest footnote

Per-country blocking is available for every country, built from the delegated statistics published by all five Regional Internet Registries. Those prefixes are only injected into the routing table once you actually enable a country: the full registry dataset runs to hundreds of thousands of prefixes, and carrying it unconditionally would cost gigabytes of router memory for data nobody asked for. It is also why we quote the threat table rather than the whole table — quoting everything would overstate what your session actually carries several times over, and contradict the number in your own dashboard.

Intelligence is aggregated from commercial and community threat feeds, malware and botnet trackers, rogue-ASN drop lists expanded into real prefixes, and the delegation files published by the five Regional Internet Registries. Refresh intervals are per-source, from fifteen minutes to hourly, with registry data daily — policy changes are near-instant, but ordinary feed data is not, and we would rather say which is which. The named source list is published in the console rather than here, pending a commercial licensing review.

12 platforms

Show, don't tell. Here is the configuration.

The console generates a complete, paste-ready configuration with your real peering address, ASN, MD5 key, next-hops and prefix limit already substituted, stamped with the categories you receive. This is what it looks like.

Cisco IOS-XE
! BlackHawk Threat BGP Route Server — AS401573
! Egress-only peering. This session never accepts a route from you.
! Transport: ebgp-multihop   Prefix limit: 250,000
! Categories: C2, MALWARE, RANSOMWARE, BULLETPROOF, ASN_DROP, PHISHING,
!             BOTNET, COMPROMISED, BRUTEFORCE, EXPLOIT,
!             SCANNER_MALICIOUS, ADMIN_EMERGENCY
!
ip route 192.0.2.1 255.255.255.255 Null0
ipv6 route 100::1/128 Null0
!
ip community-list standard BH-BLACKHOLE permit 65535:666
!
route-map BH-IN permit 10
 match community BH-BLACKHOLE
 set ip next-hop 192.0.2.1
 set ipv6 next-hop 100::1
route-map BH-IN deny 99
!
route-map BH-OUT deny 10
!
router bgp 65001
 neighbor 198.51.100.10 remote-as 401573
 neighbor 198.51.100.10 description BlackHawk BlackHole route server
 neighbor 198.51.100.10 password <md5-key-from-console>
 neighbor 198.51.100.10 update-source Loopback0
 neighbor 198.51.100.10 ebgp-multihop 16
 neighbor 2001:db8:401:573::10 remote-as 401573
 neighbor 2001:db8:401:573::10 password <md5-key-from-console>
 neighbor 2001:db8:401:573::10 update-source Loopback0
 neighbor 2001:db8:401:573::10 ebgp-multihop 16
 !
 address-family ipv4
  neighbor 198.51.100.10 activate
  neighbor 198.51.100.10 send-community both
  neighbor 198.51.100.10 route-map BH-IN in
  neighbor 198.51.100.10 route-map BH-OUT out
  neighbor 198.51.100.10 maximum-prefix 250000 90 restart 15
 exit-address-family
 !
 address-family ipv6
  neighbor 2001:db8:401:573::10 activate
  neighbor 2001:db8:401:573::10 send-community both
  neighbor 2001:db8:401:573::10 route-map BH-IN in
  neighbor 2001:db8:401:573::10 route-map BH-OUT out
  neighbor 2001:db8:401:573::10 maximum-prefix 250000 90 restart 15
 exit-address-family
!
Illustrative sample. Addresses come from the documentation ranges reserved in RFC 5737, RFC 3849 and RFC 6666 — the console substitutes your real peering address, ASN, MD5 key, next-hops and prefix limit. The templates behind these samples and the console generator are the same templates, so a syntax correction cannot land in one and not the other.

Twelve platforms, one template set

Cisco IOS-XE · Cisco IOS-XR · Cisco NX-OS · Juniper Junos · Arista EOS · Fortinet FortiOS · HPE Aruba AOS-CX · Nokia SR OS · Huawei VRP · FRRouting · VyOS · MikroTik RouterOS 7

Standards, not conventions

RFC 7999 blackhole community 65535:666 on every prefix. RFC 8092 large communities for classification, with 32-bit ASN support. RFC 2385 TCP-MD5, rotatable per router. RFC 5082 GTSM when it genuinely applies. Legacy standard communities emitted alongside, for CPE that cannot match large ones.

Your next-hop, your decision

We advertise the prefix. What your router does with it — discard to null0, or divert to a collector that watches — is entirely yours. That is also how you run this in observe mode first and see what would be dropped before anything is.

The collector — optional, on-premises

Your firewall drops the packet. It doesn't tell you which laptop sent it.

Blackholing tells you that traffic stopped. It does not tell you which machine was sending it, and that is the only fact an incident actually turns on.

Without the collector

Your router discarded 41,000 packets.

With the collector

WKS-4021 in Finance, used by Dana Ruiz, has been beaconing to a known C2 every ninety seconds since Tuesday.

The collector is a separate appliance deployed inside your own network. It sits where the blackhole routes point and turns every blocked beacon into a named internal host, a named user, and the threat category that put the route there. It is detection without deploying an agent on the endpoint — which for OT, medical devices, printers, building management, contractor laptops and anything embedded is the only way you will ever learn that box is compromised.

Sinkhole mode

The collector is the next-hop your blackhole routes point at.

Highest fidelity — every attempt, with ports, at packet granularity. Reads headers only and never stores payload. Linux only.

Flow mode

The router that discards the traffic exports NetFlow v5/v9, IPFIX or sFlow.

Nothing is redirected and the data path does not change — frequently the only proposal that survives a change board. Both sensors can run at once.

Where the names come from

  • Reverse DNS against your own internal resolvers — rate-limited and serialised, because five hundred concurrent PTR queries against a domain controller that is also serving logons is a good way for a monitoring appliance to get blamed for a slow morning.
  • Active Directory over LDAPS — two read-only searches per host yield the OU path, operating system, description, last logon and the assigned user. The bind account needs nothing beyond the read every domain already grants. Plain LDAP without StartTLS is refused at startup.
  • Logon events you already forward — Windows Security 4624 and 4768 from WEF, NXLog, Winlogbeat, Snare or your SIEM's forwarder. Only interactive logons count. Active Directory records who a machine is assigned to; only the domain controller that authenticated the session knows who was at the keyboard, and the dashboard says which source answered.

We publish our severity weights

Sorting by “most blocked attempts” puts a noisy scanner at the top and the C2 beacon on page four. One confirmed C2 destination outranks ten thousand hits on a low-risk scanner list, and here is exactly how:

Worst category severity
70 critical · 50 high · 25 medium · 10 low
Distinct destinations contacted
+20 at 20 or more · +12 at 5 · +6 at 2
Volume of attempts
+10 at 1,000 hits · +6 at 100 · +3 at 10

Nothing in this product uses machine learning, and the score is a deterministic heuristic rather than a black box. Publishing the weights is the point.

What leaves your premises, exactly

Destination addresses — the ones we already told you to blackhole — plus a heartbeat carrying version, platform and two counters. No internal addresses. No hostnames. No usernames. No flow records. Those exist only on your own appliance.

There is an automated test in the build that fails if a field is ever added to that heartbeat. A privacy promise backed by a test is a different kind of claim from a privacy promise in a PDF.

Offline mode stops all outbound calls entirely — the dashboard keeps working and classifies from the last route snapshot, so an air-gapped site is a day behind rather than unsupported. Every release artifact carries a detached minisign signature and the installer aborts on a bad one; two signing keys are trusted at once so key rotation is not a flag day. The binary runs with CAP_NET_RAW inside your network, so “verify the checksum” is not advice worth giving without something to verify against.

Inbound too

Your routes already work in both directions.

The generated configuration drops traffic your network sends to a blackholed address. Dropping traffic that arrives from one is a different mechanism — and it is one line per interface.

Because a blackholed prefix already sits in your FIB pointing at a discard next-hop, loose unicast Reverse Path Forwarding makes the question answer itself: a packet arrives from an address inside that prefix, the reverse lookup lands on a discard route, uRPF fails, and the packet is dropped in hardware — before any ACL, before the firewall, at line rate.

No new feed. No new session. No ACL to maintain. The same routes, working both ways.

Loose mode — the correct choice
interface GigabitEthernet0/0/1
 ip verify unicast source reachable-via any
 ipv6 verify unicast source reachable-via any

Use loose mode. Strict uRPF will break any multihomed network with asymmetric routing, which is most of them. This warning stays attached wherever the feature is described, including here.

Safety

A blackhole feed is dangerous. This one is built like it.

Every competitor carries this risk and most do not talk about it. If the feed is wrong, it takes a network off the air — so here is precisely what stands between a bad upstream day and your default route.

Immutable never-blackhole safelist

RFC 1918 space, loopback, link-local, CGNAT, documentation ranges, the public resolvers and the DNS root servers can never enter the table. Matched by overlap in both directions — announcing 8.8.0.0/16 is refused exactly as firmly as 8.8.8.8/32, because it would break the same resolver.

Minimum prefix length

Nothing broader than a /8 in IPv4 or a /19 in IPv6, from any source. A malformed feed cannot advertise a default route and blackhole the internet for every subscriber at once.

Per-source ceilings and shrink guards

A feed returning more than its ceiling, less than half its previous size, or nothing at all is rejected and the previous data retained. Silent truncation upstream does not become a silent gap in your protection.

Operator injections go through the same gate

Emergency injections are not a bypass. An emergency is exactly when someone pastes the wrong thing.

Test before add, on every new source

A preview fetches and parses a candidate URL and reports what it would contribute — bytes, entries parsed, unique prefixes, what the safelist stripped, and a sample of real prefixes — without writing a row to the database or a route to the table.

HTTPS-only feeds, and a global table cap

Threat data over plaintext is trivially poisoned, so plaintext sources are refused outright. A hard ceiling on the whole table bounds the blast radius of everything upstream of it, and policy changes are rate-limited to five per minute per subscriber.

Egress-only. Inbound is always rejected

One unconditional global policy statement rejects every route from every peer, and it covers a new peer the moment it connects. The service is architecturally incapable of accepting a route from you — it cannot be used to leak or hijack a prefix through your session.

A downgrade destroys nothing

Routers and categories beyond a lower plan stop being programmed, but the rows survive. An upgrade restores them with nothing re-entered. Deleting them would make "try the free tier for a month" a one-way door.

Who can do what

owner

Everything, plus billing and ownership transfer. Exactly one per organization, enforced in the database, and cannot be removed.

admin

Change routing and peering, invite and remove people. Cannot promote themselves to owner — that would be a complete takeover of the tenant, billing included.

member

Change routing and peering.

viewer

Read only. Exists so a NOC screen or an auditor can be given access without also being able to blackhole a /8.

Plans

Everything that makes a session safe is on every plan.

TCP-MD5, GTSM, prefix limits and inbound rejection are not upsells. Charging for those would be charging for correctness.

 FreeEnterprise
BGP neighbours (routers)1Unlimited
Threat categories7All
Per-country blockingEvery country
Seats3Unlimited
Prefix limit per session250,0001,000,000
TCP-MD5, GTSM, prefix limits, inbound rejectionIncludedIncluded
Multi-node peering redundancyIncludedIncluded
Emergency injectionsIncludedIncluded
IPv4 and IPv6 on the same sessionIncludedIncluded
Generated configuration, all 12 platformsIncludedIncluded
Price$0.00Priced per engagement

Everything that makes a session safe is on every plan

TCP-MD5, GTSM, prefix limits and inbound rejection are not upsells. Charging for those would be charging for correctness.

Emergency injections are free, deliberately

It is how a zero-day reaches every subscriber at once. A paywall in front of that would leave the accounts least able to respond the most exposed.

Node redundancy is free

Every router peers with every active peering node on any plan. Redundancy is never the thing being sold.

Plan enforcement lives in the routing controller, not only in the web console, so a change takes effect on the wire within about ten seconds whether or not anybody opens a page. And a downgrade destroys nothing — routers and categories beyond a lower plan stop being programmed, but the rows survive, and an upgrade restores them with nothing re-entered. Deleting them would make “try the free tier for a month” a one-way door.

The on-premises collector is entitled per plan rather than sold as a line item on this table, so how many your plan carries is settled when we talk. We do not publish an uptime figure or an SLA here, because neither is defined yet — and a number invented for a pricing page is a promise nobody agreed to.

One router config paste, and the threat stops arriving.

Sign in, enter a router address and ASN, generate a key, paste the configuration. The free tier is a working session, not a trial — and our engineers will sit on the call while you bring it up if you would rather not do it alone.

BlackHawk Threat BGP Route Server · AS401573 · Egress-only peering

Before You Ask

Questions we get every time

What is BlackHawk BlackHole?

BlackHawk BlackHole is a BGP blackhole route server operated by BlackHawk Data as AS401573. You bring up a single eBGP session from a router you already own, and that router begins discarding traffic to command-and-control servers, malware and ransomware infrastructure, phishing hosts, botnets, scanners and entire hostile networks — in the forwarding plane, at line rate, before any of it reaches a firewall rule, an IDS or a proxy licence tier. You choose which threat categories and which countries you receive, and changes apply over the live session in seconds without a reconnect.

How is this different from the firewall, IPS or secure web gateway we already have?

Different layer, different economics. Every inspection control you own examines traffic in software, and each one is priced or performance-limited by how much traffic it inspects and how many indicators it can hold — usually a few thousand. A router’s forwarding plane is the one place in the network where dropping a packet is genuinely free, and you have already paid for it. BlackHole does not inspect anything; it tells hardware you already own what to throw away. It is not a replacement for a firewall. It is the capacity already sitting idle in your routers, and it makes the firewall’s inspection budget go further by removing traffic that never needed inspecting.

Will a blackhole feed break something on our network?

That is the right question to ask, and the guardrails are built for it. An immutable safelist means RFC 1918 space, loopback, link-local, CGNAT, documentation ranges, the public resolvers and the DNS root servers can never enter the table — matched by overlap in both directions, so announcing 8.8.0.0/16 is refused exactly as firmly as 8.8.8.8/32. Nothing broader than a /8 in IPv4 or a /19 in IPv6 is accepted from any source, so a malformed feed cannot advertise a default route. Feeds that shrink by more than half, return nothing, or exceed their ceiling are rejected and the previous data retained. You also choose the next-hop, so you can start in observe mode against a collector and see exactly what would be dropped before you drop anything.

Can BlackHawk hijack or leak our routes through this session?

No. Inbound is rejected for every peer unconditionally, by a single global policy statement that covers a new peer the moment it connects. The route server sends routes and never accepts one. The service is architecturally incapable of accepting a route from a customer, which is a stronger statement than a per-peer filter that someone could misconfigure.

How long does deployment take?

Sign in, enter your router address and ASN, generate an MD5 key, and paste the configuration the console generates for your platform. It is genuinely minutes for the route server — there is no appliance, no agent and no tunnel. The optional on-premises collector is a separate, later step.

Which router platforms are supported?

Twelve, with a complete paste-ready configuration generated for each: Cisco IOS-XE, Cisco IOS-XR, Cisco NX-OS, Juniper Junos, Arista EOS, Fortinet FortiOS, HPE Aruba AOS-CX, Nokia SR OS, Huawei VRP, FRRouting, VyOS and MikroTik RouterOS 7. The templates behind the public samples and the console generator are shared, so a syntax correction cannot land in one and not the other.

Why does the generated config use ebgp-multihop rather than ttl-security?

Because GTSM (RFC 5082) requires an arriving TTL of 255, which only holds when the two BGP speakers are directly adjacent. BlackHole peers across the public internet from your loopback, so nearly every subscriber is multihop. Leaving GTSM off emits ebgp-multihop; turning it on emits ttl-security, and the console warns when the combination looks wrong. A hard-coded "ttl-security hops 1" is the single most common reason one of these sessions never leaves Active.

Does it tell us which of our machines is infected?

That is what the optional on-premises collector does, and it is the part most buyers come for. Blackholing tells you that traffic stopped; it does not tell you which machine sent it, and that is the only fact an incident actually turns on. The collector is deployed inside your own network, sits where the blackhole routes point, and turns every blocked beacon into a named internal host, a named user and the threat category that put the route there — turning "your router discarded 41,000 packets" into "WKS-4021 in Finance, used by a named employee, has been beaconing to a known C2 every ninety seconds since Tuesday." It works without deploying an agent on the endpoint, which for OT, medical devices, printers, building management and contractor hardware is the only way you will ever learn that box is compromised.

What data leaves our premises?

From the route server: your peering configuration and session telemetry. From the collector: the destination addresses your own routers already blackholed, plus a five-minute heartbeat carrying version, platform and two counters. No internal addresses, no hostnames, no usernames and no flow records — those exist only on your own appliance. There is an automated test in the build that fails if a field is ever added to that heartbeat. An offline mode stops all outbound calls entirely, so an air-gapped site is a day behind rather than unsupported.

Can we give an auditor or a NOC screen read-only access?

Yes. The viewer role is read-only and cannot change routing, which exists precisely so access can be granted without also granting the ability to blackhole a /8. Owner, admin, member and viewer are enforced in the database, and an admin cannot promote themselves to owner or remove one.

Does it support IPv6?

Full parity on the same session. MP-BGP carries IPv6 threat prefixes with correct next-hops for both families, not as an afterthought.

How much does it cost?

There is a genuinely capable free tier: one router, seven threat categories including command and control, malware, ransomware, phishing, botnets and bulletproof ranges, three seats, a 250,000-prefix session limit, and emergency injections. Enterprise adds unlimited routers and seats, every category, per-country blocking and a one-million-prefix limit, and is priced per engagement. Everything that makes a session safe — TCP-MD5, GTSM, prefix limits, inbound rejection — is on every plan, because charging for those would be charging for correctness. A downgrade destroys nothing: rows survive and an upgrade restores them with nothing re-entered.