Security of Blockchains at Capacity

Given a network of nodes with certain communication and computation capacities, what is the maximum rate at which a blockchain can run securely? We study this question for proof-of-work (PoW) and proof-of-stake (PoS) longest chain protocols under a 'bounded bandwidth' model which captures queuing and processing delays due to high block rate relative to capacity, bursty release of adversarial blocks, and in PoS, spamming due to equivocations. We demonstrate that security of both PoW and PoS longest chain, when operating at capacity, requires carefully designed scheduling policies that correctly prioritize which blocks are processed first, as we show attack strategies tailored to such policies. In PoS, we show an attack exploiting equivocations, which highlights that the throughput of the PoS longest chain protocol with a broad class of scheduling policies must decrease as the desired security error probability decreases. At the same time, through an improved analysis method, our work is the first to identify block production rates under which PoW longest chain is secure in the bounded bandwidth setting. We also present the first PoS longest chain protocol, SaPoS, which is secure with a block production rate independent of the security error probability, by using an 'equivocation removal' policy to prevent equivocation spamming.