Byzantine Fault Tolerance (BFT)

Byzantine Fault Tolerance (BFT) is a characteristic of a computer system that enables it to achieve consensus even when certain components fail.

Byzantine Fault Tolerance (BFT) is the property of a computer system that allows it to reach consensus regardless of the failure of some of its components.

What Is Byzantine Fault Tolerance (BFT)?

The Byzantine Generals’ Problem is a famous abstraction that serves to demonstrate one of the key problems in computer science, especially in regard to distributed computer systems (of which cryptocurrency is an example).

Imagine several Byzantine generals camped around an enemy city who communicate with each other only via sending messengers. They have to make a collective decision on whether to attack the city or retreat. However, some of the generals are traitors and may actively work against reaching a consensus. Is it possible to create a system that will ensure that the loyal generals decide on a common plan of action regardless of available knowledge on which generals are the traitors?

A thorough analysis of the problem shows that such an algorithm is possible, but only if more than two-thirds of generals are loyal.

Decentralized cryptocurrencies such as Bitcoin (BTC) are in essence distributed computer systems: their networks are composed of individual nodes operated by independent people or organizations that compete to process transactions and add them to the end of the blockchain. With nodes separated geographically and independent of each other or any central authority, it is impossible to reliably know which nodes are supplying faulty information about transactions — either maliciously or by accident.
Byzantine Fault Tolerance refers to the property of a distributed computer system that allows it to overcome this problem and consistently form consensus regardless of the fact that some nodes disagree with the rest, either accidentally or on purpose. This can be achieved via technical solutions, such as Bitcoin’s proof-of-work algorithm, but only if more than two-thirds of nodes remain loyal to the system.