The most fundamental service for the delivery of data on the Internet is routing, the task of computing the path that messages must follow in order to go between networks that are not directly connected. While a large body of work has studied the task of ensuring that these paths are computed correctly, surprisingly little work has been done to study the behavior, when under attack, of the system of computers, software and protocols that compute these paths, what we call the routing system.
This project is investigating the resilience of the Internet routing system to attacks that induce "churn" by forcing programs or computers to (temporarily) stop participating in the routing system. The project is devising experiments and simulations that will allow the development of mathematical models that can measure the vulnerability of the routing system to "cascading failures," caused by resource depletion, software faults, and protocol errors, that cause the entire routing system to become unavailable. The project is also developing methods to mitigate the effects of these attacks.
The project is expected to result in new data, models, and software that can be used by the community to evaluate and improve the behavior of the current routing system and proposed "future Internet" designs under destabilizing attacks, positively impacting the safety and economic health of the United States and other nations.
