Core Concepts
This paper proposes a generic Markov Decision Process (MDP) model for analyzing selfish mining attacks against a wide class of DAG-based blockchain protocols, enabling fair comparison and validation across different protocols.
Abstract
The paper presents a generic framework for modeling and analyzing selfish mining attacks against DAG-based blockchain protocols. Key points:
Existing selfish mining analyses have focused on linear chain protocols like Bitcoin, but many recent protocols use DAG structures, which require more complex MDP models.
The authors propose a modular protocol specification that captures the key components needed for selfish mining analysis, including block mining, chain update, progress tracking, and reward allocation.
Based on this protocol specification, the authors define a generic attack space and MDP model that can be instantiated for different DAG protocols. This includes tracking the defender's and attacker's views of the BlockDAG, as well as block withholding and release actions.
The authors implement their generic model for Bitcoin and validate it against previous selfish mining analyses, showing alignment with the model of Sapirshtein et al. but identifying a discrepancy with the model of Bar-Zur et al.
The authors outline plans to extend the framework to model additional security metrics beyond selfish mining revenue, such as censoring and history rewriting, as well as to handle larger BlockDAGs using reinforcement learning techniques.
The key contribution is the development of a modular and generic MDP framework that can be applied to analyze selfish mining in a wide range of DAG-based blockchain protocols, facilitating fair comparisons and validations across different designs.