|
Results:
|
Presents a formal model of BGP update propagation, which provides a
basis for defining the set of all possible outputs that can be observed at one
router when a given sequence of routing signals is input at another router.
This is then combined with a stochastic model, based on Finite Markov Chains,
for reasoning about the likelihood of a given observation.
Demonstrates that the signal input/output
relation can be extremely counterintuitive, even for small networks, which
helps to understand why the interpretation of passively collected BGP
updates is difficult.
|