Incentive Attacks on DAG-Based Blockchains with Random Transaction Selection

Several blockchain consensus protocols proposed to use Directed Acyclic Graphs
(DAGs) to solve the limited processing throughput of traditional single-chain
Proof-of-Work (PoW) blockchains. Many such protocols utilize a random transaction
selection strategy (e.g., PHANTOM, GHOSTDAG, SPECTRE, Inclusive, and Prism) to
avoid transaction duplicities across parallel blocks in DAG and thus maximize the
network throughput. However, previous research has not rigorously examined
incentive-oriented malicious behaviors when transaction selection deviates from
the protocol, which motivated our research. In the scope of this work, we perform
a game-theoretic analysis of the generic DAG-based blockchain protocol that uses
the random transaction selection strategy, proving that such a strategy does not
represent a Nash equilibrium. Furthermore, we develop a blockchain simulator that
extends existing open-source tools to support multiple chains and explore
incentive-based deviations from the protocol. Our simulations of simple network
topology with ten miners confirm our conclusion from the game-theoretic analysis.
The simulations show that malicious actors who do not follow the random
transaction selection strategy can profit more than honest miners. This has
a detrimental effect on the processing throughput of the protocol because
duplicate transactions are included in more than one block of different chains.
Moreover, we show that malicious miners are incentivized to form a shared mining
pool to increase their profit. This undermines the network's decentralization and
degrades the design of the protocols in question.