Detecting and Punishing Selfish Behavior During Gossiping in Algorand Blockchain

Abstract

Blockchains offer improved security, transparency, and anonymity for decentralized applications such as cryptocurrencies, however low efficiency and block throughput continues to be a challenge. Newer Proof-of-Stake (or PoS) systems such as Algorand provide a significantly higher block (commit) rate and throughput, but block propagation (on the peer-to-peer network) continues to remain a significant bottleneck impacting performance in such systems. One main drawback is that such systems implicitly assume that network nodes are not selfish, and that they honestly participate in propagating and validating blocks as they are broadcast or gossiped on the network. The goal of this paper is to investigate the impact of selfish behavior during block propagation (or gossip) on the security and throughput of a PoS blockchain network such as Algorand. More specifically, this paper proposes a role-based approach to detect and punish selfish nodes in Algorand. Further, a thorough game-theoretic analysis and mechanism design is conducted. Simulation experiments are done to show that the proposed detection technique can reduce selfish behavior and improve throughput in Algorand.