A BGP route leak occurs when a network propagates a route beyond the scope in which that route was intended to be advertised. The route itself may be legitimate, but an incorrect export policy can cause Internet traffic to follow an unintended path.
This distinction makes route leaks different from many other routing incidents.
The affected IP prefix may be real.
The origin Autonomous System may be legitimate.
The BGP announcement may even remain RPKI Valid.
The problem is that an intermediate Autonomous System has advertised the route to another network when routing policy says it should not have done so.
A route leak can therefore redirect traffic through infrastructure that was never intended to carry it.
The result may include:
longer network paths, congestion, dropped traffic, unstable routing or widespread service disruption.
Understanding route leaks is important because the Internet depends on thousands of independent Autonomous Systems exchanging routing information with one another.
BGP gives those operators substantial freedom to define local routing policy.
That flexibility is one of BGP's strengths.
It also means incorrect policy can propagate beyond one network.
What Is a BGP Route Leak?
The IETF defines a route leak as the propagation of routing announcements beyond their intended scope.
In practice, this often happens when a network learns a route from one neighbour and then exports it to another neighbour that should not have received it.
The technical definition and common categories are documented in RFC 7908 — Problem Definition and Classification of BGP Route Leaks.
Consider a multihomed company connected to two transit providers.
Normally:
Provider A → Customer
and:
Provider B → Customer
The customer uses both providers to reach the Internet.
But imagine the customer accidentally learns a large part of the Internet routing table from Provider A and then announces those routes to Provider B.
The network has unintentionally created:
Provider A → Customer → Provider B
The customer is now behaving like a transit network even though that was never the intended relationship.
Traffic that would normally travel directly between large transit networks may begin flowing through the customer.
That is a classic route-leak scenario.
Why Does BGP Allow This to Happen?
BGP is fundamentally a policy-based routing protocol.
It allows Autonomous Systems to decide:
- which routes to accept;
- which routes to announce;
- which neighbours to use;
- which routes to prefer;
- and how traffic should enter or leave their networks.
The protocol itself does not automatically understand every commercial or operational relationship between two networks.
A BGP session may connect:
provider ↔ customer
peer ↔ peer
or:
route server ↔ participant.
Those relationships matter because they determine which routes normally should be exported.
BGP's core protocol is defined in RFC 4271 — Border Gateway Protocol 4.
For background on Autonomous Systems and routing autonomy, see What 2026 ASN Data Tells Us About Network Growth and Internet Decentralisation — LARUS Foundation.
A Simple BGP Route Leak Example
Consider three networks:
AS64500 — Transit Provider A
AS65010 — Customer Network
AS64501 — Transit Provider B
AS65010 is multihomed.
Its intended topology is:
AS64500
↘
AS65010
↗
AS64501
AS65010 receives routes from both providers so that it can reach the Internet.
Suppose Provider A sends AS65010 a route:
203.0.113.0/24
with an AS path such as:
64500 64496
AS65010 should normally use that information internally.
It should not automatically advertise the route to Provider B.
But because of a configuration error, it sends:
65010 64500 64496
to AS64501.
Provider B may now believe that AS65010 provides a usable path toward the destination.
If that route is selected and propagated, traffic may begin travelling through the customer network.
The customer has accidentally become transit.
Why Can a Route Leak Cause an Outage?
The leaking network may not have enough infrastructure to carry the redirected traffic.
Imagine a small enterprise normally handles:
2 Gbps
of Internet traffic.
Because of a routing leak, external networks suddenly begin sending:
50 Gbps
through it.
The network may experience:
congestion
→ packet loss
→ router overload
→ service degradation
or:
complete failure.
Even if the leaking router remains online, the resulting path may be much longer or less efficient than the intended route.
RFC 7908 notes that route leaks can redirect traffic through unintended paths and may result in overload or blackholing.
BGP Route Leak vs BGP Hijack
Route leaks and route hijacks are related routing-security problems, but they are not the same.
| Routing Event | Core Problem |
|---|---|
| BGP route leak | A route is propagated beyond its intended scope |
| BGP hijack | A network originates or attracts traffic toward a prefix it is not expected to originate |
In many route leaks, the original route remains legitimate.
For example:
AS64500 → AS65010 → AS64501
The problem is not necessarily that AS65010 invented the prefix.
It may have learned a legitimate route from AS64500.
The problem is that it exported the route where it should not have done so.
With an origin hijack, the problem is often closer to:
Wrong ASN originates Prefix X.
That distinction becomes particularly important when considering RPKI.
Can RPKI Prevent BGP Route Leaks?
Not all of them.
RPKI is primarily designed to help networks validate whether an ASN is authorised to originate a prefix.
Suppose:
203.0.113.0/24
is correctly authorised for:
AS64500
The legitimate route begins:
64500
Now imagine another network learns that legitimate route and incorrectly propagates it:
65010 64500
The origin ASN is still:
64500
So Route Origin Validation may still classify the route as:
Valid.
The error exists farther along the AS path.
Therefore:
RPKI origin validation can help address incorrect route origins, but ordinary ROA validation alone does not solve every route-leak problem.
LARUS Foundation explains current RPKI adoption and Route Origin Validation in RPKI Adoption in 2026: What the Data Reveals About Routing Security.
Why Route Leaks Are About Relationships
Many BGP export rules reflect relationships between Autonomous Systems.
Three common relationships are:
Customer
A network purchases connectivity from another network.
Provider
A network provides transit to a customer.
Peer
Two networks exchange selected traffic under a lateral relationship rather than providing full transit for one another.
The expected routing policy depends heavily on these relationships.
For example, a provider may advertise broad Internet reachability to its customer.
The customer should not normally take those provider-learned routes and advertise them to another transit provider as though it were providing transit.
Likewise, routes learned from one peer are generally not supposed to be exported to another peer as free transit.
A route leak often occurs when these relationship boundaries are not correctly enforced.
What Is Valley-Free Routing?
A simplified model often used to explain Internet routing relationships is called valley-free routing.
Imagine traffic moving through economic relationships.
A path may travel:
customer → provider
then perhaps through:
provider → provider
and eventually:
provider → customer.
But certain sequences normally make little economic sense.
For example:
provider → customer → different provider
can imply that the customer is unexpectedly carrying transit traffic between two upstream networks.
Route leaks often violate these expected relationship patterns.
Real Internet relationships are more complex than a single model, but the concept helps explain why route-leak detection often needs more than prefix ownership information.
It needs an understanding of who is supposed to provide transit to whom.
Common Types of BGP Route Leaks
RFC 7908 describes several categories of route leak observed in real routing incidents.
One classic example is a multihomed customer leak, where a customer receives routes from one transit provider and announces them to another.
Another occurs when a network exports routes learned from one lateral peer to another peer.
A third can occur when peer-learned routes are exported toward a transit provider.
Internal or more-specific routes can also accidentally escape into external BGP.
These cases differ technically, but the common principle is:
Routing information escaped the scope intended by the participating networks.
Route Leaks and BGP Multihoming
Multihoming deserves particular attention because it creates multiple external routing relationships.
A multihomed network may connect to:
Provider A
and:
Provider B
and possibly:
Internet Exchange Point peers.
This gives the operator more routing flexibility and resilience.
But it also requires clear import and export policies.
A route learned from Provider A may be appropriate for the organisation's own traffic.
That does not mean the route should be exported to Provider B.
Similarly, routes learned at an IXP should not automatically be redistributed into every transit relationship.
LARUS Foundation's BGP Multihoming guide explains how multiple upstream connections create routing choice and why policy becomes central to a resilient multihomed architecture.
Route Leaks at Internet Exchange Points
Internet Exchange Points allow many independent networks to peer through shared interconnection infrastructure.
This can create efficient direct routing.
But the same environment also makes correct routing policy important.
A participant might learn routes through:
- bilateral peering;
- an IXP route server;
- private interconnections;
- and transit providers.
Each route may have a different intended export scope.
A network must therefore understand:
where the route came from
and:
where it is allowed to go next.
LARUS Foundation explains public peering and route-server infrastructure in What Is an Internet Exchange Point (IXP) and How Does It Work?.
What Are BGP Roles?
The IETF has developed mechanisms that make BGP relationships more explicit.
RFC 9234 defines BGP Roles for eBGP sessions.
The defined roles include:
Provider
Customer
Peer
Route Server
and:
Route Server Client.
The idea is that neighbouring networks can establish their expected relationship at the BGP session level.
RFC 9234 also defines mechanisms that can help identify routes that should not propagate across certain relationship boundaries.
See RFC 9234 — Route Leak Prevention and Detection Using Roles.
What Is the Only-to-Customer Attribute?
RFC 9234 introduces the Only-to-Customer, or OTC, BGP path attribute.
It is designed to help networks determine whether a route is being propagated in a way that conflicts with the expected provider-customer or peer relationship.
Conceptually, it adds information that helps answer:
Was this route intended to continue toward this type of neighbour?
This is important because classic BGP route announcements do not inherently contain a complete description of every business relationship along the path.
BGP Roles and OTC therefore add additional context for route-leak prevention and detection.
Can Prefix Filtering Prevent Route Leaks?
Filtering is one of the most important operational controls.
For a customer BGP session, a provider can maintain a list of prefixes that the customer is authorised or expected to announce.
Suppose a customer should advertise only:
203.0.113.0/24
If it suddenly begins advertising:
500,000 unrelated prefixes
the provider should not simply accept them.
A strict customer filter can reject those unexpected routes.
This is one reason accurate routing-policy data is valuable.
The IETF's operational security guidance discusses prefix filtering, AS-path filtering, max-prefix limits and other BGP controls in RFC 7454 — BGP Operations and Security.
What Is Max-Prefix Filtering?
A max-prefix limit allows a router to detect when a BGP neighbour sends far more routes than expected.
Imagine a customer normally announces:
10 prefixes.
Suddenly it announces:
900,000 prefixes.
That should be suspicious.
A max-prefix control can trigger action when the number of received routes exceeds an expected limit.
This can reduce the damage caused by some large accidental leaks.
However, max-prefix filtering is not a complete solution.
A customer expected to announce ten routes might accidentally leak only five highly attractive routes.
The route count may remain below the configured limit while the leak still causes significant problems.
Can IRR Data Help?
Internet Routing Registry information can be used to help build routing filters.
An operator may publish route objects describing expected prefix and origin relationships.
Providers can then use that data as one input when constructing customer filters.
However, IRR-based filtering depends on data quality.
If routing objects are:
outdated
missing
or:
incorrect,
automated filters may also become inaccurate.
This illustrates a recurring Internet infrastructure principle:
Automation becomes more useful when the underlying records accurately reflect operational reality.
IRR information, RPKI and live BGP observations each provide different information.
No single one should automatically be treated as a complete routing-policy source.
Why Accurate Routing Records Matter
Suppose a network legitimately announces:
203.0.113.0/24
through:
AS65010
Its provider may check several sources before accepting the route.
These can include:
contractual customer information
IRR records
RPKI
known prefix lists
and:
observed routing history.
If those systems disagree, operations become more difficult.
Accurate records therefore support better automation.
But records should describe the actual network.
They should not replace operator judgement about real routing relationships.
How Can Network Operators Reduce Route-Leak Risk?
Route-leak prevention is strongest when several controls work together.
Operators should maintain clear import and export policies for every external BGP relationship.
Customer sessions should normally be limited to routes the customer is expected to originate.
Provider-learned and peer-learned routes should not be automatically redistributed into unrelated external relationships.
Operators should also consider:
prefix filters, AS-path policies, max-prefix limits, RPKI Route Origin Validation, IRR-based filters, BGP Roles where supported, and external route monitoring.
The goal is not to rely on one control.
It is to make routing errors difficult to create and easier to detect.
Why Default Route Policies Matter
A useful operational principle is:
Do not export everything unless there is a clear reason to export everything.
A network should know what each BGP neighbour is expected to receive.
For example:
Customer
May receive broad Internet routes from a provider.
Provider
Should receive the customer's authorised prefixes—not every route the customer has learned elsewhere.
Peer
Should normally receive the network's own routes and appropriate customer routes, rather than routes learned from other peers or providers.
Explicit policy reduces the chance that a configuration change accidentally turns a non-transit network into a transit path.
Monitoring Matters Because Prevention Is Not Perfect
Even well-designed filtering can fail.
Configuration changes happen.
Routing relationships evolve.
Automation can contain bugs.
Databases can become stale.
Network operators therefore need external visibility into how their routes appear from outside their own infrastructure.
Useful monitoring can detect:
- unexpected origin ASNs;
- unusual AS paths;
- sudden route-volume changes;
- unexpected more-specific announcements;
- changes in upstream visibility;
- and routes appearing through unexpected providers.
This helps identify not only route leaks but also other routing anomalies.
Route Leaks and More-Specific Prefixes
Route leaks involving more-specific routes can be particularly disruptive because Internet routers normally use longest-prefix matching.
Suppose the Internet already sees:
203.0.112.0/20
Then an unexpected route appears for:
203.0.113.0/24
Traffic destined for the /24 will normally prefer the more-specific route.
Even if the broader /20 remains correctly announced, the leaked /24 can redirect traffic.
RFC 7908 includes accidental leakage of internal and more-specific prefixes among its documented categories.
Can a Route Leak Be Accidental?
Yes.
Many route leaks are caused by configuration errors rather than deliberate attacks.
Examples can include:
- incorrect route-map logic;
- a missing export filter;
- importing and redistributing the wrong routes;
- incorrect BGP policy automation;
- misunderstanding a peer relationship;
- or accidentally advertising internal more-specific prefixes.
This is one reason route leaks should not automatically be described as malicious hijacks.
The operational impact can be severe even when nobody intended to redirect traffic.
Can a Route Leak Be Malicious?
A routing-policy violation can theoretically be deliberate.
An attacker or dishonest network might intentionally create a path that attracts traffic.
But determining intent from BGP data alone can be difficult.
The routing system shows:
what was announced
and:
how routes propagated.
It does not necessarily reveal why the operator made the announcement.
For responsible technical analysis, it is therefore better to distinguish observable routing behaviour from assumptions about intent.
What Happens to Traffic During a Route Leak?
Several outcomes are possible.
Traffic Still Reaches the Destination
The leaking network forwards traffic onward.
The route is inefficient, but connectivity remains.
Congestion Occurs
The unexpected traffic exceeds the leaking network's capacity.
Users experience latency and packet loss.
Traffic Is Blackholed
The leaking AS has no working forwarding path toward the real destination.
Traffic is discarded.
Paths Become Unstable
Networks repeatedly select and withdraw routes as the leak propagates or is corrected.
Traffic Crosses an Unexpected Network
Even when connectivity continues, packets may traverse infrastructure that was never intended to carry them.
This can create operational and security concerns.
Why Route Leaks Are a Shared Internet Problem
A routing configuration exists inside one Autonomous System.
Its effects can extend far beyond that AS.
That happens because BGP is a distributed coordination system.
Networks make local decisions.
Those decisions can then influence neighbouring networks.
The neighbours may propagate the information farther.
This architecture allows the Internet to scale without one central routing controller.
But it creates shared responsibility.
A customer should filter its exports.
A provider should validate what it accepts.
Peers should maintain appropriate export rules.
Routing information should be monitored.
No single organisation can independently guarantee global BGP correctness.
Does Preventing Route Leaks Require Centralised Routing?
No.
Route-leak prevention does not require one organisation to decide every route on the Internet.
The stronger model is distributed validation.
Each network can apply controls at its own boundaries:
What am I willing to accept?
What am I allowed to export?
What relationship do I have with this neighbour?
Does the route match expected resource information?
Does the AS path make operational sense?
This preserves local routing autonomy while strengthening shared coordination.
That distinction fits the architecture of BGP itself.
Thousands of independent networks can make their own decisions while using common standards to reduce the risk that one mistake becomes a global incident.
BGP Route Leak Prevention Checklist
For production BGP networks, a practical review should cover:
- documented customer, provider and peer relationships;
- explicit import and export policies;
- customer prefix filtering;
- AS-path filtering where appropriate;
- max-prefix limits;
- current IRR information;
- RPKI Route Origin Validation;
- BGP Roles and OTC support where applicable;
- monitoring of external routing visibility;
- change-management review;
- and tested rollback procedures.
The most important question is not simply:
“Is the BGP session up?”
It is:
“Is this neighbour receiving only the routes it is supposed to receive?”
Frequently Asked Questions
What is a BGP route leak?
A BGP route leak occurs when routing information is propagated beyond its intended scope, violating the expected routing policy between Autonomous Systems.
What causes BGP route leaks?
Common causes include missing filters, incorrect export policies, automation mistakes, accidental redistribution and misunderstanding of provider, customer or peer relationships.
Is a route leak the same as a BGP hijack?
No. A route leak often involves a legitimate route that is propagated incorrectly. A BGP hijack commonly involves an unexpected or unauthorised route origin or route intended to attract traffic.
Can RPKI stop route leaks?
RPKI Route Origin Validation can help detect unauthorised route origins, but many route leaks preserve the legitimate origin ASN and therefore may still appear RPKI Valid.
How can networks prevent route leaks?
Useful controls include prefix filtering, AS-path policy, max-prefix limits, IRR-derived filters, RPKI validation, explicit export policies, BGP Roles and routing monitoring.
What are BGP Roles?
BGP Roles allow neighbouring eBGP networks to identify relationships such as Provider, Customer, Peer, Route Server and Route Server Client. RFC 9234 defines the mechanism.
What is the OTC attribute?
Only-to-Customer, or OTC, is a BGP path attribute defined by RFC 9234 to help detect and prevent certain route-leak scenarios.
Can a multihomed network cause a route leak?
Yes. A common route-leak scenario occurs when a multihomed customer learns routes from one provider and accidentally advertises them to another provider.
Can route leaks cause outages?
Yes. Route leaks can redirect large amounts of traffic, create congestion, cause blackholing or produce unstable and inefficient routing.
Are route leaks always malicious?
No. Many route leaks result from accidental configuration mistakes.
Does max-prefix filtering stop every route leak?
No. It can limit incidents involving unexpectedly large numbers of routes, but smaller leaks may remain below the configured limit.
Can Internet Routing Registry data help prevent leaks?
IRR data can help providers build prefix and AS filters, but its usefulness depends on the accuracy and freshness of the underlying records.
Can route leaks happen at an IXP?
Networks participating at an IXP still need appropriate routing policy. Routes learned through peers or route servers can be leaked if export controls are incorrect.
Conclusion
BGP route leaks demonstrate both the strength and the challenge of the Internet's distributed routing architecture.
BGP allows thousands of independent Autonomous Systems to decide:
who they connect to
which routes they accept
which routes they advertise
and:
which paths they prefer.
That autonomy allows the Internet to function as a network of networks.
But it also means routing policy matters.
A route can be technically legitimate and still be propagated incorrectly.
A prefix can have the correct origin ASN.
Its RPKI status can be Valid.
The destination can remain real.
And yet the route can still cause disruption because it travelled beyond the scope in which it was intended to be advertised.
That is what makes route leaks fundamentally different from a simple ownership or origin problem.
Preventing them requires more than one database or one security mechanism.
It requires several layers working together:
accurate routing policy
→ prefix filtering
→ AS relationship awareness
→ RPKI and IRR information
→ BGP Roles and route-leak controls
→ external monitoring
The goal of routing security is not to centralise route decisions, but to help independent networks coordinate more reliably while retaining control over their own routing policies.
That is the balance BGP depends on:
local autonomy supported by reliable coordination.
When those boundaries are clearly documented, technically enforced and continuously monitored, one network's routing mistake is less likely to become everyone else's outage.