Agent-based Modelling of Strategic behavior in PoW Protocols

In blockchain-based systems, such as Bitcoin’s Proof-of-Work (PoW) protocol, it is expected that a miner’s share of total block revenue is proportional to their share of the network’s total hashing power. However, a deviation to this behavior is the selfish mining strategy, an attack vector discovered by Eyal and Sirer in 2014. This strategy may lead to a miner earning more than their “fair share As a result, Bitcoin’s security assumption of an honest majority may not be sufficient. In this paper, in order to verify whether selfish mining is indeed a profitable strategy in PoW systems, we introduce an agent-based model to simulate the dynamics of selfish mining behavior. The model is by design minimalistic allowing us to analyze the effect of network latency, hashing power distribution, and network topology on relative revenue of selfish miners. We find that for high levels of latency, selfish mining is always a relatively more profitable strategy, and the results turn out to be very robust to changes in the network topology. In addition, we find that the hashing power distribution following power laws, as found empirically, can make it harder for selfish miners to be profitable. Our analysis confirms the main observations that selfish mining is always relatively more profitable for hashing powers representing more than one third of the total computing power. Further, it also confirms that selfish mining behavior could cause a statistically significant high probability of contiguously mined blocks.