The Case for Byzantine Fault Detection

Distributed systems are subject to a variety of faults and attacks. In this paper, the authors consider general (Byzantine) faults, i.e. a faulty node may exhibit arbitrary behavior. In particular, a faulty node may corrupt its local state and send arbitrary messages, including specific messages aimed at subverting the system. Many security attacks, such as censorship, freeloading, misrouting, and data corruption, can be modeled as Byzantine faults. Systems can be protected with Byzantine Fault Tolerance (BFT) techniques, which mask a bounded number of Byzantine faults, e.g. using state machine replication. BFT is a very powerful technique, but it has its costs.