When searching the internet for the terms “centralized” and “decentralized”, one quickly come up to the following figure:

This figure is originally from a publication of Paul Baran from 1962 about distributed networks. Baran describes the risks of a centralized network structure (A) and on the other side, the advantages of a distributed network structure (C). A decentralized network (B) for Baran is still a star shaped network with some kind of additional links. Since this figure is often used without taking in account the whole publication and the time and purpose it was written, a discussion about decentralization vs. distribution arose.4 This even led to contrary representations.5 In the further course of this document we will therefore only use decentralization as term.

Vitalik Buterin took up this discussion about decentralization in his blogpost6 and introduced several levels of decentralization:

• Architectural decentralization

• Political decentralization

• Logical decentralization

For Buterin “blockchains are politically decentralized (no one controls them) and architecturally decentralized (no infrastructural central point of failure) but they are logically centralized (there is one commonly agreed state and the system behaves like a single computer)".7

Centralized systems are more efficient since in a command and control structure complex agreements are omitted. On the other hand, such a system has a higher vulnerability. The whole system depends on the central unit. Decentralized systems need more communication between the nodes but are in a way fault tolerant. Decentralized systems have a long history in computer science and a well-known problem dealing with a decentralized system is the Byzantine Generals’ Problem.8 It describes the problem of a group of generals (persons or nodes) communicating with each other over one-to-one communication about an attack (common goal) while there are potential traitors among them who intentionally misleading the others. How can they come to an agreement about an attack time and what is the best strategy to achieve this goal? The problem originally comes from system theory where decentralized networks with potential communication failures experience similar problems. Derived from this problem “Byzantine Fault Tolerance is the characteristic which defines a system that tolerates the class of failures that belong to the Byzantine Generals’ Problem".9

Satoshi Nakamoto described in his paper about Bitcoin and blockchain10 one solution to this problem by using the proof-of-work algorithm. The algorithms facilitated a majority agreement in a decentralized system without a central authority. Furthermore, potentially untrustworthy parties have no chance to disrupt the system on the long run since the rules let an honest chain always win over a dishonest one.

Decentralized in the context of blockchains is usually explained by the fact that:

• nodes are locally distributed over the earth and logical distributed over the internet,

• decisions are spread over the nodes and synchronized through a consensus algorithm,

• rules all nodes obey are incorporated inherently in the software which is used by all nodes.

“A crucial part of blockchain technology is its ability to produce honesty in an anonymous environment".11 If this decentralization is betrayed the whole system becomes vulnerable to attacks. In the blockchain technology this is often expressed by the term “51% attack”. A 51% attack refers to an attack where more than 50% of the network's mining hash rate which is used to solve the proof-of-work puzzle is controlled by a single party or group. In this case this actor could block or reverse transactions and therefore double-spend coins. The costs of such an attack can be calculated.12 They rise the more nodes and calculation power is used and the more a blockchain network is decentralized by the participation of independent miners.

As it can easily be seen decentralization cannot be reduced to a purely technical view. In an ecosystem which is built by participants with commercial or power interests also social and economic factors are involved. Special interests of groups or nations might undermine decentralization. At the same time decentralization of a blockchain network is a flowing process. Even if a high degree of distribution is reached this can change over time and therefore has to be monitored constantly.