Decentralized security mechanisms for routing protocols

Today's Internet routing protocols are built upon the basic incorrect assumption that routers propagate truthful routing information. As a result, the entire Internet infrastructure is vulnerable to security attacks from routers that propagate incorrect routing information. In fact, a single router is capable of hijacking a significant fraction of routes by launching such an attack. This issue is not just restricted to Internet routing protocols but is widely prevalent in several routing protocols that have been proposed in the research literature. Many existing approaches for addressing the security problems of routing protocols typically assume the existence of a Public-Key Infrastructure (PKI) or some form of prior key distribution mechanism along with a central authority. While a PKI does enable addressing this security threat, building one such key-distribution infrastructure may not always be feasible. One faces serious deployment barriers in building an Internet-wide PKI with a central authority especially given that deploying one such architecture requires approval across political and economic boundaries. Previous efforts for securing Internet routing and the Domain Name System using a PKI have not moved towards adoption. In this dissertation, we address the following question: Using purely decentralized mechanisms (void of a PKI and a central authority), what is the best level of security achievable for a routing protocol in the presence of adversaries? One of the key conclusions that we arrive at is the direct relationship between decentralized security and the reliable communication problem. The reliable communication problem relates to determining the constraints under which a set of good nodes in a network can reliably communicate messages between themselves in the face of adversarial nodes in the network. We show theoretical results on the constraints under which the reliable communication problem is solvable. Based on these results, we describe the design of a reliable communication toolkit that implements our algorithms and provides a suite of generic security primitives that can be used to secure a variety of routing protocols. These security mechanisms supported by the toolkit are well suited for Internet routing since they are both easy to deploy as well as offer good security guarantees. We also show that the toolkit has broader applicability beyond routing protocols to: (a) achieve decentralized key distribution; (b) address the data integrity threat to the Domain Name System (DNS) in a decentralized manner.