This project investigates the design of peer-to-peer networks which are both provably scalable and attack-resistant. The networks are scalable in the sense that time, space and bandwidth resource costs for all major operations are polylogarithmic in the number of nodes in the network. The networks are attack-resistant in the sense that most of the services are guaranteed to most of the participants, even after massive adversarial attack. The approach of this project is primarily theoretical, i.e. algorithm design and analysis. However, the project also includes building prototype systems to empirically test the algorithms.
This project has potential for significant impact in other areas of research. One major deliverable of the project is algorithms for creating a p2p network which, even under attack, will enable sharing of massive data sets. This type of network could have an impact on several scientific disciplines, including, physics (e.g. for projects like GriPhyN), astronomy (e.g. for projects like Sloan Digital Sky Survey), and biology (e.g. for projects like the Human Genome Project). Another major deliverable of this project is algorithms for creating a p2p network which, even under attack, will enable distributed computation. This type of network would allow scientific projects to tap into idle desktop computers to efficiently and robustly do lengthy computations. Astronomy (e.g. for projects like SETI@home) and biology (e.g. for projects like FOLDING@home) are examples of disciplines that would benefit from such a network. Finally, it is expected that the project will provide theoretical insights into the problem of building robust embedded networks, which share many similarities with p2p networks.
