In an earlier post I argued why Bitcoin’s stability is fundamentally a game-theoretic proposition, and ended with some questions:
Can we effectively model the system with all its interacting components in the language of strategies and payoff-maximization? Is the resulting model tractable — can we analyze it mathematically or using simulations? And most importantly, do its predictions match what we observe in practice?
Let’s look at those questions in the context of a “block withholding attack” between mining pools.
Recall that mining pools are groups of individual miners who pool their computing power as well as their rewards. Suppose two mining pools — let’s call them blue and red — are both seeking to maximize their mining rewards. Let’s say the manager of the red pool decides to infiltrate the blue pool and decrease their efficiency using some of the mining power that red (directly or indirectly) controls. This can be done by submitting shares (partial proofs of work) to earn a share of rewards, but withholding any valid blocks which are found and therefore not contributing any productive work to the blue pool. At first sight this seems like cutting off your nose to spite your face — sure, blue’s efficiency will be hurt, but red is wasting hash power as well.