Reliable Routing Protocols in Networks with Byzantine Failures

The network layer in a computer network is responsible for providing a datagram service, in which a packet issued from a source node to a destination node is expected to reach the destination node in a reasonably short period of time with some probability. The datagram service does not provide delivery guarantees: these are expected to be provided by a higher level transport layer, on top of a basic datagram service. But in order to design a reasonably efficient transport layer on top of a datagram service, the datagram service should provide the advertised service with high probability. The network layer functionality is typically provided by routers connected by links. These routers implement a protocol, which can be considered to be a distributed algorithm: each router performs the algorithm locally, operating on local data stored in local data structures, and the overall result is a global pattern of behavior. Many widely used routing algorithms are robust in the presence of simple failures (e.g. a router crashes and stops working completely, or a link goes down). However, a router with a Byzantine failure may act arbitrarily or even maliciously. We outline two network layer protocols that can route successfully in the presence of Byzantine failures as described in [2].