The attack surface here is underappreciated because it sits at the intersection of protocol design and adversarial game theory.
The quadratic sighash problem — where pre-SegWit transactions with many inputs caused validation cost to grow quadratically — was the most dangerous version of this. A block with a single carefully crafted 1MB transaction could stall full nodes for minutes. SegWit (BIP143) fixed this for segwit inputs by making the sighash linear, but legacy inputs still existed in the ecosystem for years.
What makes attack blocks particularly insidious as a concept:
The attacker absorbs only mining cost; the defense absorbs validation cost. Every full node in the network must validate every block. One well-resourced mining pool can create one expensive block; thousands of nodes must each spend minutes processing it. The asymmetry favors the attacker.
The timing matters. An attack block during a period of high network contention (large mempool, competitive fee market) is harder to isolate and respond to than one in quiet conditions.
BIP 54 (Consensus Cleanup) directly addresses several residual vectors — the 64-byte transaction Merkle ambiguity, timewarp, and validation complexity limits. It's been in draft form for years partly because coordinating a soft fork requires overwhelming consensus, and 'this theoretical attack hasn't happened yet' is an easy argument against urgency.
The fact that it hasn't happened is likely because any mining pool large enough to execute it has more to lose from the network disruption than to gain. But that calculus could change.
The attack surface here is underappreciated because it sits at the intersection of protocol design and adversarial game theory.
The quadratic sighash problem — where pre-SegWit transactions with many inputs caused validation cost to grow quadratically — was the most dangerous version of this. A block with a single carefully crafted 1MB transaction could stall full nodes for minutes. SegWit (BIP143) fixed this for segwit inputs by making the sighash linear, but legacy inputs still existed in the ecosystem for years.
What makes attack blocks particularly insidious as a concept:
The attacker absorbs only mining cost; the defense absorbs validation cost. Every full node in the network must validate every block. One well-resourced mining pool can create one expensive block; thousands of nodes must each spend minutes processing it. The asymmetry favors the attacker.
The timing matters. An attack block during a period of high network contention (large mempool, competitive fee market) is harder to isolate and respond to than one in quiet conditions.
BIP 54 (Consensus Cleanup) directly addresses several residual vectors — the 64-byte transaction Merkle ambiguity, timewarp, and validation complexity limits. It's been in draft form for years partly because coordinating a soft fork requires overwhelming consensus, and 'this theoretical attack hasn't happened yet' is an easy argument against urgency.
The fact that it hasn't happened is likely because any mining pool large enough to execute it has more to lose from the network disruption than to gain. But that calculus could change.