What BGP Actually Does
The internet is a collection of independent networks called Autonomous Systems (ASes). Your ISP is one. AWS is one. Cloudflare is one. Each AS has a unique number (an ASN) and owns blocks of IP addresses.
BGP — the Border Gateway Protocol — is how these networks tell each other how to reach those IP address blocks. When your traffic needs to get from a browser in Berlin to a server in Singapore, BGP is what builds the map of AS hops it takes to get there.
Every AS advertises prefixes (IP ranges) it can route traffic to. Neighboring ASes receive those advertisements, prepend their own ASN, and pass them along. The result is a global, distributed routing table that every major network maintains independently.
Critically: BGP is built on trust. There is no cryptographic verification of whether an AS actually has the right to advertise a given prefix.
What a Route Leak Is
A route leak happens when an AS advertises prefixes it shouldn't — typically re-advertising routes it learned from one peer out to another peer or provider without authorization.
A common scenario:
- Your network connects to both ISP A and ISP B.
- ISP A sends you their full routing table.
- Due to a misconfiguration, your router re-advertises ISP A's routes to ISP B.
- ISP B, seeing what looks like a valid path, starts sending traffic through you.
- You're now acting as a transit provider you never agreed to be — and you probably can't handle it.
The damage spreads because BGP prefers more-specific or shorter AS-path routes. If a leaked route looks more attractive than the legitimate one, routers across the internet will start preferring it. Traffic that should go directly to, say, an AWS data center starts detouring through a small regional ISP with a 1 Gbps uplink.
Why It Causes Outages
Route leaks cause outages in a few distinct ways:
- Congestion at the leaking AS: Traffic volumes the AS was never designed for saturate their links. Packets are dropped. Services time out.
- Longer, higher-latency paths: Even if traffic gets through, RTT spikes from milliseconds to hundreds of milliseconds, breaking latency-sensitive protocols and TLS handshake timeouts.
- Black-holing: If the leaking AS can't actually forward the traffic, packets simply disappear. No ICMP unreachable, no TCP RST — just silence.
- Partial reachability: Because BGP propagation is uneven, some parts of the world see the leaked route and some don't. Your service looks up from Japan but down from Europe.
Partial reachability is particularly painful to debug. Your own monitoring, running from a single region, may show green while a significant portion of real users can't connect.
Real-World Examples
The 2010 China Telecom incident involved ~37,000 prefixes being rerouted through China for about 18 minutes. In 2019, a small Pennsylvania ISP leaked over 200,000 routes through Verizon (who had no route filters in place), disrupting Cloudflare, Amazon, and others. Neither was malicious — both were misconfigurations.
How to Detect It
From your own infrastructure, route leaks are nearly invisible until users start complaining. Useful tools:
- BGPmon / Kentik — commercial BGP monitoring that alerts on unexpected prefix advertisements
- RIPE RIS or RouteViews — public route collectors you can query to see how your prefixes look globally
bgp.he.net— quick sanity check on what AS paths exist to your prefixes right now- Traceroutes from multiple vantage points — RIPE Atlas lets you run traceroutes from hundreds of global probes
This is also where multi-region uptime monitoring earns its keep. If your checks are running from five or six geographically distributed locations, a partial black-hole shows up immediately as a regional failure pattern — even before you know to look at BGP. A single-region check will lie to you.
Mitigations (For Network Operators)
If you run your own ASN:
- Register your prefixes in an IRR (Internet Routing Registry) and keep records current
- Implement RPKI (Resource Public Key Infrastructure) — cryptographically signs which AS is authorized to originate your prefixes
- Set strict BGP filters on your peers and customers; don't accept routes more specific than /24 for IPv4
- Use the
no-exportcommunity on routes you receive from transit providers to prevent accidental re-advertisement - Limit prefix counts from peers — if a customer suddenly advertises 50,000 routes, something is wrong
RPKI adoption has grown substantially in recent years and is the most meaningful systemic fix. Networks that have deployed RPKI-based Route Origin Validation will automatically reject invalid route origins.
Key Takeaways
- BGP routes traffic between autonomous systems based on trust, not cryptographic verification
- A route leak is a misconfigured re-advertisement of prefixes an AS has no business forwarding
- Leaks cause congestion, latency spikes, and black-holes — often asymmetrically across regions
- Your service can be unreachable to half the world while your single-region monitoring shows all green
- RPKI is the strongest available mitigation; IRR hygiene and strict peer filters are table stakes
- Public tools like RIPE Atlas and RouteViews let you see your routes from the outside in