Jessica Yu ANS Communication Inc. Feb. 9th, 1998 A Routing Filtering Model for Improving Global Routing Robustness - an IOPS proposal Jessica Yu ANS Communication Inc. Feb. 9th, 1998
Background Internet suffered a series of outages as a result of leakage of ‘bad’ routing information IOPS identifies a mechanism to improve routing system robustness IOPS encourages its member and other ISPs to implement it
What’s the Problem? Current global routing system is open thus vulnerable ‘Bad’ routing information injecting from anywhere of the Internet will be propagated allover resulting outages (one dead mouse spoils the whole pot of soup) Proven by several incidents occurred - unfortunately
What’s the Problem? (con’t) Could happen due to misconfiguration, software bugs or malicious attempt The ‘openness’ can be potentially used by attackers to inject routes for denial of service attacks such as smurf attack
Current Picture: Vulnerable
Bad routing information can cause big outage
IOPS Goals To identify or define mechanism to improve global routing robustness Using existing technology as much as possible so it can be implemented fast since the problem is an urgent one
The Routing Filtering Model Sparse filtering on ‘trusted’ and Dense filtering on ‘less-trusted’ routing boundaries Trusted routing boundary - the peer does dense filtering at all of its customers boundaries and all of it’s downstream providers implement such routing filtering model Otherwise, it’s less-trusted boundary This is mainly verify against the border AS which advertising the route
Scenario 1: All ISPs Dense Filter customers
Scenario 2: Some ISPs do not dense filter customers, immediate upstreams have to
Scenario 3: ISP does not dense filter less-trusted downstream ISPs, peer have to
Benefits bad routes will be stopped from propagating near or at its source thus reduce impact to a small scope of the Internet Attackers has less chance to introduce routes on the fly to the whole Internet and launch attacks To localize the impact of the ‘bad’ routes To Reducing the weakness which potentially will be explored by attackers The more ISPs or AS implement this model the better protection we have The more ISP does this, the less work everyone need to do ( less boundaries needs dense filtering, and shorter prefix each has to manage
Tools The key to this is to maintain an prefix list to generate filtering list on the border routers Private database which install customer to be announced prefix IRR Other means
Issues how to validate if someone is entitled to advertise a route Provider check its customers route match with address assignment information Add features in IRR to validate such information in combination with assignment information (working under way)
Future work This work was done 6 months ago which at the time was the best viable option This is model of checking against neighbor AS, another model is to check against the Origin AS If there is other viable proposal which will address the problem better, will consider that.