Bitcoin under Volatile Block Rewards: How Mempool Statistics Can Influence Bitcoin Mining

TL;DR

This paper develops a reinforcement learning tool to analyze Bitcoin mining strategies under volatile block rewards influenced by mempool behavior, revealing new security risks and strategic incentives as rewards diminish.

Contribution

It introduces a realistic volatile reward model using A3C reinforcement learning to evaluate mining strategies and security thresholds in Bitcoin's evolving incentive landscape.

Findings

Adversarial strategies like selfish mining can be more profitable post-reward halving.

The implicit predictability of transaction arrivals underestimates security thresholds.

Transition to fee-based rewards incentivizes pools to prioritize immediate profits.

Abstract

The security of Bitcoin protocols is deeply dependent on the incentives provided to miners, which come from a combination of block rewards and transaction fees. As Bitcoin experiences more halving events, the protocol reward converges to zero, making transaction fees the primary source of miner rewards. This shift in Bitcoin's incentivization mechanism, which introduces volatility into block rewards, leads to the emergence of new security threats or intensifies existing ones. Previous security analyses of Bitcoin have either considered a fixed block reward model or a highly simplified volatile model, overlooking the complexities of Bitcoin's mempool behavior. This paper presents a reinforcement learning-based tool to develop mining strategies under a more realistic volatile model. We employ the Asynchronous Advantage Actor-Critic (A3C) algorithm, which efficiently handles dynamic environments, such as the Bitcoin mempool, to derive near-optimal mining strategies when interacting with an environment that models the complexity of the Bitcoin mempool. This tool enables the analysis of adversarial mining strategies, such as selfish mining and undercutting, both before and after difficulty adjustments, providing insights into the effects of mining attacks in both the short and long term. We revisit the Bitcoin security threshold presented in the WeRLman paper and demonstrate that the implicit predictability of valuable transaction arrivals in this model leads to an underestimation of the reported threshold. Additionally, we show that, while adversarial strategies like selfish mining under the fixed reward model incur an initial loss period of at least two weeks, the transition toward a transaction-fee era incentivizes mining pools to abandon honest mining for immediate profits. This incentive is expected to become more significant as the protocol reward approaches zero in the future.