Evolving Bitcoin Custody

The broad topic of this thesis is the design and analysis of Bitcoin custody systems. Both the technology and threat landscape are evolving constantly. Therefore, custody systems, defence strategies, and risk models should be adaptive too. We introduce Bitcoin custody by describing the different types, design principles, phases and functions of custody systems. We review the technology stack of these systems and focus on the fundamentals; key-management and privacy. We present a perspective we call the systems view. It is an attempt to capture the full complexity of a custody system, including technology, people, and processes. We review existing custody systems and standards. We explore Bitcoin covenants. This is a mechanism to enforce constraints on transaction sequences. Although previous work has proposed how to construct and apply Bitcoin covenants, these require modifying the consensus rules of Bitcoin, a notoriously difficult task. We introduce the first detailed exposition and security analysis of a deleted-key covenant protocol, which is compatible with current consensus rules. We demonstrate a range of security models for deleted-key covenants which seem practical, in particular, when applied in autonomous (user-controlled) custody systems. We conclude with a comparative analysis with previous proposals. Covenants are often proclaimed to be an important primitive for custody systems, but no complete design has been proposed to validate that claim. To address this, we propose an autonomous custody system called Ajolote which uses deleted-key covenants to enforce a vault sequence. We evaluate Ajolote with; a model of its state dynamics, a privacy analysis, and a risk model. We propose a threat model for custody systems which captures a realistic attacker for a system with offline devices and user-verification. We perform ceremony analysis to construct the risk model.