Levels of Decentralization and Trust in Cryptocurrencies: Consensus, Governance and Applications

Since the apparition of Bitcoin, decentralization has become an ideal praised almost religiously. Indeed, removing the need for a central authority prevents many forms of abuse that could be performed by a trusted third party, especially when there are no transparency and accountability mechanisms in place. Decentralization is however a very subtle concept that has limits. In this thesis, we look at the decentralization of blockchains at three different levels. First we look at the consensus protocol, which is the heart of any decentralized system. The Nakamoto protocol, used by Bitcoin, has been shown to induce centralization through the shift to mining pools. Additionally, it is heavily criticized for the enormous amount of energy it requires. We propose a protocol, Fantômette, that incorporates incentives at its core and that consumes much less energy than Bitcoin and other proof-of-work based cryptocurrencies. If the consensus protocol makes it possible to decentralize the enforcement of rules in a cryptocurrency, there is still the question of who decides on the rules. Indeed, if a central authority is able to determine what those rules are then the fact that they are enforced in a decentralized way does not make it a decentralized system. We study the governance structure of Bitcoin and Ethereum by making measurements of their GitHub repositories and providing quantitative ways to compare their level of centralization by using appropriate metrics based on centrality measures. Finally, many applications are now built on top of blockchains. These can also induce or straightforwardly lead to centralization, for example by requiring that users register their identities to comply with regulations. We show how identities can be registered on blockchains in a decentralized and privacy-preserving way