BGP borked? Blame the net's big boppers
Researcher says routes are leaking because ISP giants aren't filtering route info
The Internet's fundamental routing infrastructure, the Border Gateway Protocol (BGP), is so fragile that errors in one to four per cent of ISP route filters can propagate bad routes.
So says Czech DDOS-defender Qrator Labs, which carried out a project to try and help the Internet community quantify how much of the threat to BGP arises from simple things like router misconfiguration.
The company's Alexander Azimov explained the research in this post at the Asia-Pacific Network Information Centre (APNIC).
The conclusion: you don't need many mistakes to allow BGP hijacks and route leaks to have a serious impact.
Qrator Labs focussed on how well service providers implement their most common mitigation technique: ingress filters built on Internet Routing Registry (IRR) databases.
“The idea is simple: using route objects and AS-SETs to create ingress filters at customers links”, Azimov explained.
AS-SETs are a group of autonomous systems with the same routing policies.
BGP hijacks and leaks already hint that “some ISPs do not set filters at all, or there are vastly malformed AS-SETs”, he continued.
Measurement, however, isn't too difficult a challenge: if a route leak is received from a customer, but it does not belong to the “customer cone” (the Autonomous Systems reachable only through customer links), then there's a malformed IRR filter.
“We learned that seven per cent of transit ISPs in IPv4 and one per cent of transit ISPs in IPv6 are accepting leaks outside of their customer cone”, Azimov wrote – and although small, that's enough to spread bad data far and wide.
The main reason: acceptance of leaks is heavily distributed towards Tier-1 providers, making it certain that route leaks will propagate. Azimov said in the Qrator test, all Tier-1 providers accepted leaks, and half of the global top 400 ISPs handling IPv4 traffic.
Things are better in the IPv6 world – because there are fewer networks and ISPs, and therefore fewer routing anomalies.
Two ISPs, DE-CIX and Rostelecom, told Azimov they'll check their filters and work with customers on the issue.
Azimov's advice is the same as BGP admins should have been following from time immemorial, but it's worth restating: keep AS-SETs (AS object sets) and route objects current; put IRR filters on all customer links; watch for customers adding “something weird” in their AS-SETs; consider applying IRR filters to peering links; and “constantly monitor BGP”.
Azimov presented his study at Apricot 2018, in the video below. ®
Sponsored: Beyond the Data Frontier