Dynamic Practical Byzantine Fault Tolerance

This paper describes a novel Byzantine fault tolerant protocol that allows replicas to join and exit dynamically. With the astonishing success of cryptocurrencies, people attach great importance in “blockchain” and robust Byzantine fault tolerant (BFT) protocols for consensus. Among the conventional wisdom, the Practical Byzantine Fault Tolerance (PBFT), proposed by Miguel and Liskov in 1999, occupies an important position. Although PBFT has many advantages, it has fatal disadvantages. Firstly, it works in a completely enclosed environment, where users who want to add or take out any node must stop the whole system. Secondly, although PBFT guarantees liveness and safety if at most [n-1/ 3] c out of a total n replicas are faulty, it takes no measure to deal with these ineffective or malicious replicas, which is harmful to the system and will cause system crash finally. These drawbacks are unbearable in practice. In order to solve them, we present an alternative, Dynamic PBFT.