Selfish Mining Re-Examined

Abstract

Six years after the introduction of selfish mining, its counterintuitive findings continue to create confusion. In this paper, we comprehensively address one particular source of misunderstandings, related to difficulty adjustments. We first present a novel, modified selfish mining strategy, called intermittent selfish mining, that, perplexingly, is more profitable than honest mining even when the attacker performs no selfish mining after a difficulty adjustment. Simulations show that even in the most conservative scenario (\(\gamma =0\)), an intermittent selfish miner above 37% hash power earns more coins per time unit than their fair share. We then broadly examine the profitability of selfish mining under several difficulty adjustment algorithms (DAAs) used in popular cryptocurrencies. We present a taxonomy of popular difficulty adjustment algorithms, quantify the effects of algorithmic choices on hash fluctuations, and show how resistant different DAA families are to selfish mining.